Dual-mode network interface controller

Abstract

This disclosure relates to a dual-mode network interface controller, an apparatus and method for a hybrid network interface controller (NIC), integrating an optical communication interface and a wireless communication interface, and including intelligent control circuitry for dynamically selecting or concurrently utilizing one or both interfaces based on real-time operating conditions. The NIC includes a controller and, in some embodiments, also includes a machine learning model configured to evaluate workload information and communication medium performance data to generate configuration instructions. These instructions are used to activate, deactivate, or modify the behavior of the optical and wireless interfaces, thereby optimizing link selection, performance, or energy efficiency. The hybrid NIC can operate autonomously based on local thresholds, utilize model-driven decision-making, or accept configuration commands from an external orchestrator. This architecture supports dynamic, stream-based, or system-wide medium selection, enhancing network adaptability, reliability, and performance in heterogeneous or high-demand computing environments.

Description

Technical Field

[0001] This disclosure relates generally to networking, and in particular to networking devices and methods of operation thereof. Background Technology

[0002] High-performance computing environments, cloud data centers, and edge computing infrastructures typically rely on increasingly sophisticated network interface technologies to meet the ever-increasing demands for bandwidth, latency sensitivity, and diverse workloads. Traditional network interface controllers (NICs) can be equipped with wired communication interfaces, such as Ethernet or optical transceivers, providing reliable and high-throughput data delivery for a wide range of applications. Wireless communication technologies have also made significant progress, offering increased data rates, reduced latency, and improved reliability through standards such as Wi-Fi 6 / 6E, 5G, and emerging millimeter-wave protocols. These capabilities have enabled wireless communication to serve as a viable complement to wired network interconnects in certain performance-critical and flexible deployment scenarios.

[0003] Despite these advancements, modern network workloads continue to exhibit variability in terms of bandwidth requirements, latency constraints, mobility, and availability expectations. Therefore, system designers must frequently evaluate trade-offs between different communication interfaces and manage transitions or failover between them, potentially introducing complexity and inefficiencies. Consequently, improvements are needed in how network architectures can manage communication and support the changing demands of computing and interconnected systems. Summary of the Invention

[0004] According to at least some aspects of this disclosure, methods and systems relate to a hybrid network interface controller (NIC) that integrates both optical and wireless communication interfaces and includes intelligent control circuitry for dynamically selecting or concurrently utilizing one or both interfaces based on real-time operating conditions. This enables data centers to seamlessly select between optical and wireless links without requiring manual reconfiguration or complex network management. In some aspects, determining when to shift traffic to the most efficient link (e.g., an optical or wired link) can be based on monitored link conditions (e.g., signal strength and bandwidth usage) and workload demands. Therefore, data centers can adapt to changing workloads and high-priority demands, reducing downtime and ensuring more efficient bandwidth utilization in large-scale deployments. The NIC includes a controller and, in some embodiments, a machine learning model configured to evaluate workload information and communication medium performance data to generate configuration instructions. These instructions are used to activate, deactivate, or modify the behavior of the optical and wireless interfaces to optimize link selection, performance, or energy efficiency. The hybrid NIC can operate autonomously based on local thresholds, utilize model-driven decision-making, or accept configuration commands from an external orchestrator. This architecture enables dynamic, stream-based, or system-wide media selection, enhancing network adaptability, reliability, and performance in heterogeneous or high-demand computing environments.

[0005] An exemplary aspect of this disclosure provides a network interface device. In some aspects, the network interface device includes: an optical interface for transmitting data via an optical communication medium; a wireless interface for transmitting data via a wireless communication medium; and one or more circuits for selectively routing data from a host device to at least one of the optical interface or the wireless interface in response to a threshold condition.

[0006] The aspects include, wherein the one or more circuits are further configured to: switch data transmission from a first communication medium to a second communication medium; or concurrently enable both communication media.

[0007] The aspects include, wherein the one or more circuits further route data via the optical interface and the wireless interface in response to the threshold condition.

[0008] The aspects include, wherein the threshold condition is based on at least one of the following: a measured communication link quality metric associated with at least one of the optical communication medium or the wireless communication medium; a measurement of bandwidth utilization associated with at least one of the optical communication medium or the wireless communication medium; a detected bit error rate associated with at least one of the optical communication medium or the wireless communication medium; or a signal-to-noise ratio associated with at least one of the optical communication medium or the wireless communication medium.

[0009] The aspects include, wherein the one or more circuits are further configured to: receive an indication of workload demand; and compare the indication of workload demand with the threshold condition.

[0010] The aspects include, wherein the one or more circuits are further configured to: route high-priority workload traffic via the optical interface; and route lower-priority workload traffic via the wireless interface.

[0011] The aspects include, wherein the one or more circuits include local memory storing AI-based decision logic, and the threshold condition is determined by the AI-based decision logic in response to usage metrics collected from the optical interface and the wireless interface.

[0012] The aspects include, wherein the threshold condition is triggered in response to receiving a switching command from an external orchestration controller in the data center network.

[0013] The aspects include, wherein the network interface device may include an interface for coupling the network interface device to the host device, wherein the interface is further configured to communicate with the host device using at least one of an optical path or an electrical path.

[0014] The aspects include, wherein the wireless interface includes at least one of the following: a photonic mixer stage for generating optical signals; or a radio frequency mixer stage configured to generate or receive millimeter wave or terahertz signals.

[0015] The aspects include, wherein the network interface device may include a photonic signal generator for generating a modulated optical signal; and a branching module for directing the modulated optical signal to at least one of the optical interface or the wireless interface.

[0016] The aspects include, wherein the network interface device may include an optical output that couples the modulated optical signal to an optical fiber.

[0017] The aspects include, wherein the wireless interface comprises: a photonic mixer stage for generating millimeter-wave or terahertz signals; and an antenna coupled to the photonic mixer stage.

[0018] The aspects include, wherein the controller is used to adaptively adjust the operation of the network interface device based on at least one of real-time network conditions or workload requirements, the adjustment of the network interface device including at least one of: modifying data routing logic, selecting between communication media, or reallocating bandwidth on the optical interface and the wireless interface.

[0019] An exemplary aspect of this disclosure provides a network interface device. In some aspects, the network interface device includes: an optical interface for receiving a modulated optical signal; optical processing circuitry for demodulating the modulated optical signal; a wireless interface for receiving a wireless signal; wireless processing circuitry for decoding the wireless signal; and one or more output circuits coupled to the optical processing circuitry and the wireless processing circuitry, the one or more output circuits being used to provide demodulated data to a host system.

[0020] The aspects include, wherein the network interface device includes a controller communicatively coupled to the optical processing circuitry and the wireless processing circuitry, the controller being configured to: monitor at least one link quality metric associated with the optical interface and the wireless interface; and selectively enable or disable reception through at least one of the optical interface or the wireless interface based on threshold conditions.

[0021] The aspects include, wherein the controller is also configured to receive data concurrently through the optical interface and the wireless interface in response to detecting that at least one of the total bandwidth requirement or link utilization meets or exceeds the threshold condition.

[0022] The aspects include, wherein the controller includes AI-based decision logic for modifying the threshold conditions based on the at least one link-quality metric.

[0023] An example aspect of this disclosure provides a system. In some aspects, the system includes: a host device for generating outgoing data traffic; a network interface device communicatively coupled to the host device, the network interface device including: an optical interface for transmitting data via an optical communication medium; a wireless interface for transmitting data via a wireless communication medium; and a controller for selectively routing the outgoing data traffic between the optical interface and the wireless interface based on threshold conditions.

[0024] The network interface device further includes: an optical processing circuit for demodulating incoming optical signals from the optical communication medium; and a wireless processing circuit for decoding incoming wireless signals from the wireless communication medium, wherein the controller is further configured to route incoming data to the host device.

[0025] The aspects include, wherein the controller includes an artificial intelligence module for: monitoring link performance metrics from the optical interface and the wireless interface; and dynamically adjusting the threshold conditions in response to changes in the link performance metrics.

[0026] The aspects include, wherein the host device includes a workload manager that assigns priority indicators to outgoing data, and the controller is configured to route higher priority traffic through the optical interface and lower priority traffic through the wireless interface.

[0027] An exemplary aspect of this disclosure provides a hybrid network switch. In some aspects, the hybrid network switch includes: a plurality of wired ports for transmitting and receiving data via physical cabling; a plurality of wireless ports for transmitting and receiving data via wireless communication media; and switching hardware coupled to the plurality of wired ports and the plurality of wireless ports, the switching hardware being configured to route data from an ingress port to a selected egress port based on routing or switching logic.

[0028] This document describes additional features and advantages, which will be apparent from the following detailed description and accompanying drawings. Attached Figure Description

[0029] This disclosure has been described in conjunction with the accompanying drawings, which are not necessarily drawn to scale.

[0030] Figure 1A A block diagram of a hybrid network interface controller having an optical interface and a wireless interface according to at least one aspect of the present disclosure is depicted.

[0031] Figure 1B A block diagram of another hybrid network interface controller having an optical interface and a wireless interface according to at least one aspect of this disclosure is depicted.

[0032] Figure 2A A block diagram of a controller having a model-based communication medium selector according to at least one aspect of this disclosure is depicted.

[0033] Figure 2B A block diagram is depicted of a controller for a communication medium selector with external configuration according to at least one aspect of this disclosure.

[0034] Figure 2C A block diagram of an external configuration generation system for controlling interface selection, according to at least one aspect of this disclosure, is depicted.

[0035] Figure 2D A data structure according to at least one aspect of this disclosure is described.

[0036] Figure 3A A block diagram of a training system for generating machine learning models according to at least one aspect of this disclosure is depicted.

[0037] Figure 3B A block diagram depicting a neural network architecture according to at least one aspect of this disclosure is provided.

[0038] Figure 4A A block diagram depicts an optical module having transmitting and receiving paths according to at least one aspect of this disclosure.

[0039] Figure 4B A block diagram depicts a wireless module having transmit and receive paths according to at least one aspect of this disclosure.

[0040] Figure 4C A block diagram depicts a photonics-based wireless module that uses optical mixing to generate signals according to at least one aspect of this disclosure.

[0041] Figure 4D A block diagram of a hybrid optical transmission system with selective output routing according to at least one aspect of this disclosure is depicted.

[0042] Figure 4E A block diagram of a photonic mixer stage according to at least one aspect of this disclosure is depicted.

[0043] Figure 5A A flowchart is depicted for a method of dynamically selecting or modifying an active communication interface according to at least one aspect of this disclosure.

[0044] Figure 5B A flowchart is depicted for a method for generating communication interface configuration output using a decision model, according to at least one aspect of this disclosure.

[0045] Figure 5C A flowchart is depicted for a method of applying configuration output to a hybrid network interface controller according to at least one aspect of this disclosure.

[0046] Figure 6 A flowchart is depicted illustrating an exemplary method for dynamically selecting between an optical communication interface and a wireless communication interface according to at least one aspect of this disclosure.

[0047] Figure 7A The figure shows a block diagram of a system including a switching circuit located between a first group of devices and a second group of devices according to at least one embodiment.

[0048] Figure 7B The figure shows a block diagram of a hybrid network switch according to aspects of this disclosure.

[0049] Figure 8 The figure shows a block diagram of an example system according to at least one embodiment.

[0050] Figure 9 The figure illustrates an example data center according to at least one embodiment.

[0051] Figure 10The figure illustrates a processing system according to at least one embodiment.

[0052] Figure 11 The figure illustrates a computer system according to at least one embodiment.

[0053] Figure 12 The figure illustrates a system according to at least one embodiment.

[0054] Figure 13 The figure illustrates an example integrated circuit according to at least one embodiment.

[0055] Figure 14 The figure illustrates a computing system according to at least one embodiment.

[0056] In the various figures, the same reference numerals and names indicate the same elements. Detailed Implementation

[0057] The following description provides only examples and is not intended to limit the scope, applicability, or configuration of the claims. Rather, the following description will provide those skilled in the art with an enabling description of how to implement the described embodiments. It should be understood that various changes may be made to the function and arrangement of the elements without departing from the spirit and scope of the appended claims.

[0058] As will be understood from the following description, for computational efficiency reasons, system components can be placed at any suitable location within a distributed network of components without affecting the operation of the system.

[0059] Furthermore, it should be understood that the various links connecting the elements can be wired, traced, or wireless links, or any suitable combination thereof, or any other suitable known or later-developed element capable of supplying and / or transmitting data to and from the connected element. For example, the transmission medium used as a link can be any suitable carrier for electrical signals, including coaxial cables, copper wires and optical fibers, electrical traces on printed circuit boards (PCBs), etc.

[0060] As used herein, the term "automatic" and its variations refer to any appropriate process or operation performed without substantial human input. However, a process or operation can be automatic even if substantial or non-substantial human input is used in its execution, provided that input is received prior to its execution. Human input is considered substantial if it affects how the process or operation will be performed. Human input that consents to the execution of a process or operation should not be considered "substantial."

[0061] As used herein, the terms “determine,” “calculate,” and “operate,” and their variations thereof, are used interchangeably and include any appropriate type of method, process, operation, or technique.

[0062] Various aspects of this disclosure will be described herein with reference to the accompanying drawings, which are schematic illustrations of an idealized configuration. Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. It will also be understood that terms such as those defined in common dictionaries shall be interpreted as having a meaning consistent with their meaning in the context of the relevant art and this disclosure.

[0063] This disclosure relates to apparatus, methods, processing systems, and computer-readable media for dynamically selecting among multiple communication media in a network interface device. In some aspects, data centers and high-performance computing environments can utilize flexible network interconnection solutions that adapt to different workloads and changing network conditions. Aspects of this disclosure describe a dual-mode hybrid NIC that integrates an optical communication interface and a wireless communication interface with control circuitry that selects the communication medium based on real-time conditions and requirements. In some aspects, network interface devices may be limited to a single communication medium, making them unsuitable for adapting to changing network conditions, congestion, or varying workload demands. This flexibility can lead to poor performance when the primary medium experiences degradation, bandwidth limitations, or latency issues. Furthermore, conventional network architectures often require manual reconfiguration or complex orchestration to switch between different communication media, resulting in operational inefficiencies and potential service interruptions.

[0064] Various aspects of this disclosure address these challenges by providing a hybrid network interface device, which may be identical or similar to a hybrid network interface controller, combining an optical communication interface and a wireless communication interface controlled by selection circuitry. In some aspects, the hybrid architecture provides automatic dynamic switching between communication media (e.g., optical (fiber) and wireless connections) based on real-time network conditions, real-time performance metrics, workload characteristics and demands, and resource utilization, thereby ensuring seamless, high-performance connectivity in the data center. In other aspects, an artificial intelligence (AI) module monitors link quality metrics, bandwidth requirements, and workload priorities, and adjusts threshold conditions to modify routing performed by the hybrid network interface device. The hybrid network interface device can operate in multiple modes, including single-media operation, concurrent dual-media operation for load balancing, or priority-based traffic routing across different media.

[0065] One technical solution offers several advantages over conventional methods. For example, aspects of this disclosure improve network resilience by automatically rerouting traffic when a communication medium experiences degradation, thereby preventing service interruptions and maintaining connectivity. As another example, aspects of this disclosure can improve overall system efficiency by optimizing resource utilization through traffic allocation across available media based on workload characteristics and current network conditions. As yet another example, AI-based decision logic can be used to improve optimization performance based on historical patterns and learned relationships between network conditions and application requirements. As yet another example, hybrid network interface devices enhance deployment flexibility in diverse environments where different communication media can have their own advantages, allowing the system to select the optimal path for each specific workload or operating condition.

[0066] Figure 1A An example of a network interface device 100A according to at least one aspect of this disclosure is depicted, the network interface device 100A being configured to communicate with a host system 156 using multiple communication media. The network interface device 100A can be implemented via a hybrid network interface controller, which is a pluggable module, expansion card, or embedded interface device in a computing system. The hybrid network interface controller 100A enables dynamic selection between optical communication channels and wireless communication channels, thereby providing flexibility, redundancy, and performance adaptability based on system conditions or operational requirements.

[0067] In some aspects, the hybrid network interface controller 100A includes an optical interface 104 and a wireless interface 108. In some aspects, the optical interface 104 may include an optical module 116 and an optical connector interface 120. The optical module 116 may include circuitry configured to transmit and receive optical signals via an optical communication medium such as an optical fiber. The optical module 116 may include various optical components, including but not limited to one or more laser diodes for signal generation, a photodetector for optical signal reception, an optical modulator for encoding data onto the optical signal, a demodulator for recovering data from the received optical signal, and signal conditioning components for ensuring signal integrity. The optical module 116 may be configured to support various optical transmission standards and protocols, including but not limited to Ethernet, Fibre Channel, InfiniBand, or proprietary protocols. In some implementations, the optical module 116 may incorporate integrated photonic circuitry, silicon photonics, or other advanced optical technologies to minimize size, power consumption, or latency. The optical interface 104 may include diagnostic capabilities for monitoring optical power levels, wavelength stability, or bit error rate.

[0068] The optical connector interface 120 can be configured to couple the optical module 116 to an external optical cable. The optical connector interface 120 can be compatible with industry-standard optical connectors, such as LC (Lucent connector), SC (Subscriber connector), MPO (Multi-Fiber Push-In), or other suitable optical connectors. In some implementations, the optical module 116 can be designed as an industry-standard pluggable transceiver module, such as SFP (Small Form Factor Pluggable), QSFP (Quad Small Form Factor Pluggable), or OSFP (Eight Small Form Factor Pluggable).

[0069] In some aspects, the wireless interface 108 includes a wireless module 124 and a wireless connector interface 128. The wireless module 124 may include circuitry configured to transmit and receive wireless signals via a wireless communication medium. Wireless communication can utilize various frequency bands and protocols, including but not limited to millimeter waves (e.g., 30-300 GHz), terahertz (e.g., 0.1-10 THz), or other radio frequency bands suitable for high-bandwidth, short-range communication. The wireless module 124 may include one or more radio frequency (RF) front-end components, such as power amplifiers, low-noise amplifiers, mixers, filters, and / or oscillators. Such components can process the RF signals used in wireless communication. The wireless module 124 may also include a baseband processor that handles digital signal processing functions such as modulation, demodulation, encoding, decoding, error correction, and / or protocol processing. The wireless module 124 may also incorporate media access control (MAC) logic to manage channel access, addressing, and frame formatting according to the wireless communication protocol.

[0070] The wireless connector interface 128 can be configured to couple the wireless module 124 to one or more antenna elements 129 or waveguides. In some aspects, the antenna elements 129 or waveguides may be located inside the card package and / or include external components connected via suitable RF connectors. The antenna element 129 can be configured for directional transmission to enhance signal strength in a specific direction or for omnidirectional coverage for broader connectivity. In some aspects, the wireless interface 108 can utilize beamforming techniques with phased array antennas to direct wireless signals toward one or more target receivers to improve signal quality and reduce interference.

[0071] In some aspects, controller 112 may be communicatively coupled to optical interface 104 and wireless interface 108. Controller 112 may include logic configured to route data traffic to optical module 116 or wireless module 124 based on communication policies, performance metrics, or channel conditions. In some implementations, controller 112 resides at least partially within local memory 111. As used herein, local memory may refer to one or more memory components located on or integrated with the network interface controller hardware, which may store executable firmware, logical states, decision rules, or configuration parameters for real-time operation. In some aspects, controller 112 may execute or access operational instructions directly from local memory 111, thereby providing access to link monitoring data, thresholds, or routing logic without interacting with host system memory or an external orchestrator. In some aspects, controller 112 includes or communicates with a communication medium selector 140, which is configured to select one or both of available communication paths (e.g., optical and / or wireless). The communication medium selector 140 can incorporate decision logic for evaluating one or more operating parameters to determine the optimal communication medium based on current conditions. In some aspects, operating parameters may include, but are not limited to, bandwidth requirements, latency constraints, error rate, signal strength, power consumption targets, or application-specific quality of service requirements.

[0072] like Figure 1A As further illustrated, controller 112 may include selectors 144 and 148, or be operatively associated with them. Selector 144 corresponds to an optical communication path, and selector 148 corresponds to a wireless communication path. Selectors 144 and 148 may be implemented through various mechanisms, including but not limited to electrical switches, software-defined logic gates, and / or routing rules, which determine which interface is actively used to send / receive data to / from host system 156 and / or which interface is actively used to send / receive data to / from another communication device.

[0073] In some implementations, selector 144 and selector 148 can operate mutually exclusively, thereby activating optical interface 104 or wireless interface 108. In other implementations, selector 144 and selector 148 can be activated simultaneously to achieve concurrent data transmission or reception through optical interface 104 and wireless interface 108. This concurrent operation can support load balancing, data redundancy, or channel bonding to increase aggregate bandwidth.

[0074] In some aspects, coupling interface 132 is configured to physically and electrically couple the hybrid network interface controller 100A to the host system 156. Coupling interface 132 may provide a communication path between the network interface controller and the host system 156, which is installed with and / or otherwise communicates with it. In some aspects, coupling interface 132 may include a Peripheral Component Fast Interconnect (PCIe) interface, an edge connector, or any suitable bus connector compatible with a host system interface such as host system interface 136. Coupling interface 132 may support various communication protocols and standards for host integration, including but not limited to PCIe, CXL (Compute Fast Link), or proprietary interconnect technologies. In some aspects, coupling interface 132 is configured to exchange bidirectional data and control signals with host device 156. Coupling interface 132 may include a PCIe Gen4 x8 connection or other suitable interconnect supporting memory read / write operations and configuration commands. Figure 1A The illustration depicts the logical communication between the NIC and host device 156, enabling workload-aware routing and real-time performance adaptation.

[0075] In some aspects, host system 156 represents a computing platform that uses hybrid network interface controller 100A for network communication. In some aspects, host system 156 may be a server, storage system, network interconnection device, or other computing device utilizing network connectivity. In some aspects, host system 156 communicates with hybrid network interface controller 100A via coupling interface 132, sending outbound data for transmission and receiving inbound data from the network connection.

[0076] Figure 1B The figure illustrates another example of a hybrid network interface controller 100B according to at least some aspects of this disclosure. The hybrid network interface controller 100B may include an optical interface 104 and a wireless interface 108, each communicatively coupled to a controller 112. The hybrid network interface controller 100B may be electrically coupled and physically coupled to a host system 156 via a coupling interface 132. In some aspects, the optical interface 104 may be associated with a MAC circuit 152, which is communicatively coupled to both the optical module 116 and the controller 112. In some aspects, the MAC circuit 152 performs data link layer functions for the optical communication path. The MAC circuit 152 may be implemented as a dedicated digital logic block, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or firmware executed by a microcontroller or processor. In some aspects, the MAC circuit 152 performs functions including Ethernet or other protocol framing, packet encapsulation, error detection and correction, flow control, and media arbitration. In some respects, MAC circuitry 152 provides an abstraction layer between the physical layer functions performed by optical module 116 and higher-level protocol processing that may occur in controller 112 or host system 156.

[0077] In some respects, the wireless module 124A can be associated with the wireless MAC circuit 154. The wireless MAC circuit 154 can implement protocol-specific logic for various wireless communication standards, including Wi-Fi (IEEE 802.11 family), 5G NR (New Radio), millimeter-wave protocols, or other wireless technologies suitable for high-speed data transmission. The wireless MAC circuit 154 can support wireless-specific tasks such as frame generation and processing, retransmission management, acknowledgment processing, encryption and security protocols, and wireless channel access coordination.

[0078] Now for reference Figure 2A An example of controller 112A is illustrated according to at least one aspect of this disclosure. Controller 112A can be implemented as previously described. Figure 1A and Figure 1B This is a part of the described controller 112. Controller 112A can be configured to select about Figure 1A-1B The described hybrid network interface controller 100A and / or 100B includes one or more communication interfaces. In some aspects, controller 112A may include a communication media selector 202A. The communication media selector 202A can be configured to control or influence data traffic routing to two or more communication interfaces. The communication media selector 202A can be implemented using digital logic circuitry, a programmable microcontroller, an ASIC, an FPGA, and / or combinations thereof. In some aspects, the communication media selector 202A includes decision logic for evaluating input parameters and generating control signals to activate, deactivate, or modify the behavior of the communication interfaces. The communication media selector 202A can operate on configuration information to determine which communication path is enabled for a given data transmission.

[0079] exist Figure 2A In the described aspects, the communication medium selector 202A can control the selection of one or more of an optical communication path via optical selector 144 and a wireless communication path via wireless selector 148. Selectors 144 and 148 can be used as previously described... Figure 1A-1B The aforementioned switching circuits, multiplexers, logic gates, or other routing control mechanisms are used for implementation. In some aspects, selector 144 and / or selector 148 may provide gating functions that enable or disable signal transmission through their respective communication paths. When activated by communication medium selector 202A, optical selector 144 allows data transmission via the optical interface, thereby enabling communication via the optical communication medium. Similarly, when activated by communication medium selector 202A, wireless selector 148 allows data transmission via the wireless interface, thereby enabling communication via the wireless communication medium.

[0080] The operation of the communication medium selector 202A can be controlled at least in part by configuration information 214. Configuration information 214 can instruct the communication medium selector 202A how it should operate (e.g., whether to select an optical communication interface, a wireless communication interface, or to select both interfaces concurrently (e.g., in load balancing or failover mode)). Configuration information 214 may include selection parameters, thresholds, timing constraints, and / or operating modes indicating which communication interface should be active under various conditions. In some aspects, configuration information 214 may include interface priority specifications, traffic routing rules, and / or bandwidth allocation instructions that control how data is routed between available communication media.

[0081] In some aspects, configuration information 214 can be generated by model 204. Model 204 may include logic, including but not limited to rule-based engines, finite state machines, statistical estimators, or trained machine learning models, for evaluating communication conditions and / or workload profiles. In some implementations, model 204 may be a neural network, decision tree, support vector machine, or other trained machine learning algorithm for predicting the optimal communication interface selection based on observed network conditions, application requirements, and / or system states. Model 204 may combine deterministic logic (e.g., threshold-based rules) and / or probabilistic reasoning (e.g., statistical reasoning) to generate interface selection decisions under different operational scenarios.

[0082] In some aspects, model 204 can be configured to receive one or more inputs, including workload information 210 and / or communication medium information 206, and generate configuration information 214 in response to receiving such information. In some aspects, model 204 can analyze patterns or trends in the input data to predict which communication interface will provide optimal performance for a given workload and / or network conditions. Model 204 can apply weighting factors to different input parameters based on the correlation between different input parameters and current operational objectives (e.g., maximizing throughput, minimizing latency, reducing power consumption).

[0083] Workload information 210 may reflect characteristics of the current application workload, such as, but not limited to, bandwidth requirements, latency sensitivity, data type, priority level, or protocol requirements. For example, workload information 210 may indicate that the current data stream requires sustained high bandwidth with moderate latency sensitivity, suggesting a preference for an optical interface. Alternatively, workload information 210 may indicate a surge transmission pattern with strict latency constraints, which may favor a wireless interface. In some aspects, workload information 210 may include application identifiers, service level indicators, or priority labels that help model 204 determine the relative importance of different performance metrics for the current data transmission.

[0084] A workload source can provide workload information 210. In some aspects, a workload source can refer to an external component, such as an orchestration engine, scheduler, or hypervisor, that communicates with the controller 112A via a control plane interface. A workload source can analyze application requirements, monitor system resources, or trace service level objectives to generate workload characterizations. In some implementations, a workload source can pre-classify traffic flows based on application profiles, thereby assigning priority levels or performance requirements that make interface selection decisions informed.

[0085] Communication medium information 206 may include operational parameters or telemetry data describing the performance, quality, or status of one or more communication interfaces. For example, communication medium information 206 may include signal-to-noise ratio (SNR), link latency, throughput, jitter, packet error rate, or energy consumption metrics. These metrics may be measured in real time or aggregated over various time intervals to provide both immediate and historical perspectives on interface performance. In some aspects, communication medium information 206 includes link operation status indicators, congestion notifications, or stability assessments reflecting the current operating conditions of each available communication medium.

[0086] Communication medium information 206 may originate from one or more external sources, such as communication medium information source 208 (e.g., optical transceiver, wireless chipset, or monitoring engine). Communication medium information source 208 may include sensor circuitry, counters, timers, or analysis components for measuring and reporting relevant performance metrics. In some implementations, communication medium information source 208 includes diagnostic logic for detecting anomalies, predicting degradation trends, or identifying optimal operating parameters for each communication interface.

[0087] In some aspects, communication medium information 206 may be derived at least in part from control plane information 216, such as one or more metrics or policies delivered from a management system or orchestrator. Control plane information 216 may include network-wide state updates, congestion notifications, or global optimization instructions that influence interface selection decisions. In some aspects, control plane information 216 provides contextual data that helps model 204 interpret local measurements relevant to broader system conditions or objectives. In some implementations, communication medium information 206 and / or configuration information 214 may be provided via control plane information 216.

[0088] In some implementations, configuration information 214 can be received directly from an external source 212 (e.g., a management system or orchestrator), completely bypassing model 204. For example, the orchestration system can provide static or policy-driven configuration commands that instruct the communication medium selector 202A to operate in a predefined mode (e.g., "optical only", "wireless preferred", or "failover enabled"). This direct configuration path can be utilized in scenarios where centralized control is preferred in local decision-making or where system-wide policies must cover interface-specific performance considerations.

[0089] In some aspects, configuration information 214 may optionally be provided to the communication medium selector 202A via control plane information path 216. Control plane information path 216 may represent a dedicated communication channel, message queue, register interface, or API through which configuration instructions are delivered. Depending on the implementation, control plane information path 216 may also serve as input to model 204 or be used to provide communication medium information 206.

[0090] The communication medium selector 202A can communicate with other components of the controller 112A and / or other systems within the hybrid network interface controller via path 218. Path 218 can represent one or more data or control buses, interconnects, or messaging interfaces used to deliver selection results, receive feedback, or update operational status. In some aspects, path 218 carries control signals to activate or deactivate interface components, adjust transmission parameters, or reconfigure data paths based on the selected communication medium.

[0091] Now for reference Figure 2B Another example of controller 112B is described according to at least one aspect of this disclosure. Controller 112B can be configured to manage communication path selection in a hybrid network interface controller using configuration logic from an external source. Controller 112B can be implemented as previously described in Figure 1A and Figure 1B The specific configuration of controller 112 described in the document.

[0092] In some aspects, controller 112B may include communication medium selector 202B, which may be the same as or similar to communication medium selector 202A. Figure 2ASimilarly, the selection output of the communication medium selector 202B can be provided via optical selector 144 and wireless selector 148. Each of the optical selector 144 and wireless selector 148 can be implemented as a control signal, a switchable routing path, and / or an interface enable mechanism, configured to activate and / or deactivate the corresponding communication medium. When activated, optical selector 144 enables data transmission via the optical interface; similarly, when activated, wireless selector 148 enables data transmission via the wireless interface. In some implementations, both selectors can be active simultaneously to support concurrent use of multiple communication paths.

[0093] and Figure 2A In contrast to embodiments where the local model 204 can generate configuration information, aspects of the controller 112B can be configured to receive configuration information 214 from one or more external sources. The configuration information 214 can determine the operational state of the communication medium selector 202B (e.g., whether to select an optical interface, a wireless interface, or both). The configuration information 214 may include selection commands, operating parameters, or mode settings for control interface activation and data routing decisions.

[0094] Configuration information 214 can be received via control plane information path 216 and can originate from various sources. Control plane information path 216 can represent a management interface, control channel, or management API through which external systems can influence or direct the behavior of controller 112B. In some respects, control plane information path 216 supports bidirectional communication, thereby allowing controller 112B to report status information or performance metrics to configuration sources.

[0095] For example, configuration information 214 can be derived from or delivered directly by model 204. Model 204 can reside outside of controller 112B and can be implemented as a decision engine or machine learning inference model that operates on workloads or link layer data. By positioning model 204 externally, various aspects of controller 112B can utilize centralized decision-making, potentially leveraging more comprehensive system knowledge or computational resources than are available locally. External model 204 can incorporate network-wide policy considerations, global resource optimization objectives, or cross-device coordination constraints that are difficult to capture within a device's local decision-making framework.

[0096] Model 204 can generate configuration information 214 based on communication medium information 206 (e.g., telemetry from a transceiver or monitoring agent) and / or workload information 210 (e.g., throughput requirements, latency requirements). In some implementations, model 204 can be applied within controller 112B to implement potentially impractical analysis or learning algorithms. This can include ensemble models, deep learning networks, or reinforcement learning methods that adapt interface selection strategies based on observations across multiple devices or time periods.

[0097] As previously described, communication medium information 206 may reflect current or historical performance metrics associated with the communication medium, including but not limited to signal strength, congestion level, packet loss, jitter, or link availability. Communication medium information 206 may originate from internal monitoring systems or external components and be delivered as part of the control plane interface. In some aspects, communication medium information 206 may include aggregated statistics, trend analyses, or predictive metrics that help anticipate future performance changes or degradation patterns.

[0098] As previously described, workload information 210 can describe application-level or system-level data requirements, including service class, priority level, protocol characteristics, or power constraints. In some implementations, workload information 210 includes application profiles, traffic patterns, or resource utilization metrics that characterize the current operating context. Workload information 210 can originate from a workload source, which can be an orchestrator, scheduling agent, or system controller external to controller 112B. The workload source can analyze application requirements, monitor system resources, or enforce service level protocols to generate appropriate interface selection instructions.

[0099] In another aspect, model 204 may reside in an external management system and / or orchestrator 220. The management system and / or orchestrator 220 may include an orchestrator, a software-defined networking (SDN) controller, or a management policy engine. The management system and / or orchestrator 220 may specify selection behavior based on global policies, system-wide load balancing, or failover logic. In some implementations, the management system and / or orchestrator 220 provides high-level instructions that complement or override the model-generated recommendations and workload-specific configurations.

[0100] In some respects, the communication medium selector 202B can communicate with other components of the controller 112B and / or other systems within the hybrid network interface controller via path 218. Path 218 can represent one or more data or control buses, interconnects, or messaging interfaces used to deliver selection results, receive feedback, or update operational status. In some respects, path 218 carries control signals to activate or deactivate interface components, adjust transmission parameters, or reconfigure data paths based on the selected communication medium.

[0101] Now for reference Figure 2C An external management system and / or orchestrator 220 is illustrated according to at least one aspect of this disclosure. The external management system and / or orchestrator 220 can be configured to generate configuration information for controlling the selection of communication media in one or more hybrid network interface controllers. In some aspects, the external management system and / or orchestrator 220 can be implemented externally to the controller 112 and / or hybrid network interface controller 100A / 100B depicted in the previous figures (e.g., within an orchestrator operating in a distributed network environment, a control plane manager, a network management system, or a dedicated configuration engine).

[0102] External management system and / or orchestrator 220 may include model 204, which can be associated with... Figure 2A and Figure 2B The model 204 described herein is identical or similar. Model 204 in the external management system and / or orchestrator 220 can be configured to receive and process multiple input signals and generate one or more instances of configuration information 214. In some aspects, model 204 can be implemented using various forms of decision logic, including but not limited to rule-based decision engines, policy evaluation frameworks, finite state machines, statistical analyzers, and / or machine learning models trained on network conditions and workload patterns as previously described. As previously described, when implemented as a machine learning model, model 204 can incorporate neural networks, decision trees, reinforcement learning algorithms, or other computational methods that can identify patterns in the input data and generate optimal configuration instructions based on those patterns.

[0103] In some respects, model 204 can receive communication medium information 206, which can communicate with... Figure 2A and Figure 2BThe described communication medium information is the same or similar. Communication medium information 206 can be collected from various sources throughout the network, including from the hybrid network interface controller itself, from network switches or routers, from dedicated monitoring probes, or from a management system that aggregates performance metrics across multiple devices. In some aspects, communication medium information 206 can include not only raw performance metrics but also derived analyses, such as link stability predictions, bandwidth predictions, or anomaly detection results. These derived analyses can provide contextual information that helps model 204 make more informed configuration decisions anticipating future network conditions, rather than simply reacting to the current state.

[0104] Model 204 can also receive workload information 210, which can be used in conjunction with information about... Figure 2A and Figure 2B The workload information 210 described is the same as or similar to that described above. In the context of an external management system and / or orchestrator 220, workload information 210 may describe the characteristics or requirements of one or more applications, services, or tenants operating within a network-connected environment. Workload information 210 may include parameters such as latency sensitivity (e.g., real-time vs. bulk transmission requirements), throughput requirements (e.g., surge rate, sustained bandwidth), security constraints (e.g., encryption requirements, traffic isolation), or job prioritization information (e.g., critical vs. background tasks). In some aspects, workload information 210 may be provided by a variety of workload-aware systems, including but not limited to application orchestrators, virtual machine monitors, container management platforms, cloud scheduling agents, or similar systems with visibility into application requirements and behavioral characteristics. Such systems may generate workload profiles characterizing the network requirements of different application components and allow model 204 to select communication media based on specific workload requirements. In some implementations, workload information 210 may include not only current requirement information but also predictive metrics of future workload patterns, thereby enabling anticipatory configuration adjustments for anticipated changes in application behavior.

[0105] Using these inputs (communication medium information 206 and workload information 210), model 204 can generate one or more sets of configuration information 214, which can be used in conjunction with information about... Figure 2A and Figure 2B The configuration information 214 described herein is the same as or similar to that described in the previous section. Configuration information 214 may instruct downstream communication medium selectors (e.g., such as...). Figure 2A or Figure 2BThe configuration information 214 may indicate how the optical communication path should be preferred for certain types of traffic, whether the wireless interface should be used as the primary or secondary path for a specific data flow, or whether both media should be activated concurrently to support load balancing or redundancy requirements.

[0106] In some respects, configuration information 214 includes not only basic selection instructions but also additional control metadata, such as activation thresholds (e.g., signal quality levels that trigger media switching), policy priorities (e.g., energy efficiency versus maximum performance), timing parameters (e.g., hysteresis settings to prevent rapid oscillations between interfaces), or conditions that should trigger re-evaluation (e.g., significant changes in workload characteristics or media performance). This control metadata allows the communication media selector in the hybrid network interface controller to make decisions that consider multiple operational objectives and / or adapt to changing conditions without requiring continuous reconfiguration from an external management system and / or orchestrator 220.

[0107] In some respects, Figure 2C The architecture described represents a configuration generation logic separated from the controller itself, allowing for global or multi-device coordination of communication policies. This separation can provide centralized optimization of network resources across multiple devices, coordination of interface selection decisions (to avoid potential RF interference or optical congestion), and / or consistent policy enforcement across the entire network domain. In some implementations, this approach can facilitate network management paradigms, including but not limited to intent-based networking, software-defined networking (SDN) orchestration, and / or AI-driven network adaptation that optimizes communication medium selection based on changing and / or current operating conditions and service requirements.

[0108] In some respects, the external management system and / or orchestrator 220 can generate configuration information 214 for multiple hybrid network interface controllers, thereby enabling coordinated decisions that take into account system-wide resource constraints and communication optimization opportunities. The external management system and / or orchestrator 220 can also integrate with a broader management framework, allowing communication media selection to be managed by high-level business policies, service level agreements, or operational priorities established at the organizational level.

[0109] Now for reference Figure 2D According to at least one aspect of this disclosure, methods for combining Figure 2A-2C Example data structures in the described communication medium selection framework. Figure 2D It describes how communication medium information 206, configuration information 214 and workload information 210 are each represented or implemented as a set of parameters related to the communication path and / or an operational standard associated with the hybrid network interface controller.

[0110] In some respects, communication medium information 206 can be related to... Figure 2A , Figure 2B and Figure 2C The described communication medium information 206 is the same as or similar to that described in the previous section. Communication medium information 206 represents structured data describing one or more communication interfaces available in a hybrid network interface system. In some aspects, communication medium information 206 may include data from... Figure 1A and Figure 1B The optical interface 104 and wireless interface 108 described herein collect or are associated with various real-time and / or historical telemetry values ​​related to interface performance.

[0111] Among other parameters, communication medium information 206 may include wireless link parameters 222 and optical link parameters 224. Wireless link parameters 222 may include multiple measurements or metrics characterizing the performance, status, or quality of the wireless communication link. In some implementations, wireless link parameters 222 may include, but are not limited to: Received Signal Strength Indication (RSSI), which quantifies the power present in the received radio signal; SNR, which measures the ratio of signal power to noise power; wireless delay measurement, which indicates the time delay of packet transmission in milliseconds; jitter metric, which describes the variation in packet delay; link error rate, which quantifies the frequency of transmission errors; and / or channel utilization metric, which indicates the percentage of available bandwidth currently in use.

[0112] Wireless link parameters 222 can be obtained through various measurement techniques, including direct sampling of radio frequency signals, statistical analysis of packet transmission characteristics, or other methods. Figure 1A The polling of hardware registers within the wireless module 124 described herein. In some respects, wireless link parameters 222 can be collected at different time scales, ranging from instantaneous readings to time averages reflecting sustained performance trends. These parameters can be represented in various formats, including raw values, normalized metrics, category classifications (e.g., “excellent,” “good,” “average,” “poor”), or metrics indicating whether performance improves, deteriorates, or stabilizes over time.

[0113] Optical link parameters 224 may include multiple measurements or metrics characterizing the performance, status, or quality of the optical communication link. In some implementations, optical link parameters 224 may include, but are not limited to: optical power level, which measures the strength of the received or transmitted optical signal in decibels (dBm); forward error correction (FEC) rate, which indicates how frequently error correction mechanisms are used to recover from signal degradation; bit error rate (BER), which quantifies the proportion of erroneous bits received relative to the total number of bits transmitted; dispersion measure, which characterizes the spread of optical pulses during transmission; and / or optical signal-to-noise ratio (OSNR), which measures the ratio of signal power to noise power in the optical domain.

[0114] Optical link parameters 224 can be collected through various measurement techniques, including direct optical power measurement, statistical analysis of received digital errors, or monitoring of optical signals. Figure 1A The optical transceiver diagnostic data for optical module 116 described herein. Similar to wireless link parameters 222, optical link parameters 224 can be collected at different time scales and represented in various formats to provide real-time operational status and long-term performance trends that can enable informed decisions on communication medium selection.

[0115] Configuration information 214 can be related to About Figure 2A , 2B The configuration information 214 described in 2C is the same as or similar to that described in 2C. Configuration information 214 may represent information used to control the communication medium selector (e.g., in...). Figure 2A and Figure 2B The decision output of the behavior of the communication medium selector 202A or 202B described herein. In some aspects, configuration information 214 includes executable instructions that determine how data will be routed through the available communication interfaces of the hybrid network interface controller. Figure 2D As depicted, configuration information 214 may include optical configuration parameters 226 and wireless configuration parameters 228. Optical configuration parameters 226 may define operational settings or selection instructions related to the optical interface, such as interface enable flags (e.g., binary indicators specifying whether the optical interface should be active or inactive), priority level assignments (e.g., numerical or classification values ​​indicating a preference level for routing certain traffic types via the optical medium), time scheduling instructions (e.g., time patterns for interface activation or deactivation), and / or failover conditions (e.g., trigger thresholds determining when traffic should be moved to an alternative interface). Optical configuration parameters 226 may include quality of service mappings that associate different traffic classifications with specific optical delivery behaviors or resource allocations.

[0116] Similarly, wireless configuration parameters 228 may include settings related to the behavior of the wireless interface, such as preferred frequency bands (e.g., channel selection instructions), modulation schemes (e.g., instructions regarding signal coding techniques), antenna control instructions (e.g., beamforming instructions), and / or throughput thresholds (e.g., minimum bandwidth requirements that must be met for the wireless interface to be considered feasible). Wireless configuration parameters 228 may include power management instructions that control the transmit power level, duty cycle behavior, or power-saving mode of the wireless interface.

[0117] In some implementations, both optical configuration parameter 226 and wireless configuration parameter 228 can include selection logic that relates to the conditions under which each corresponding interface should have exclusive use in a load-sharing or primary / backup configuration. Such selection logic can be represented as conditional statements, decision trees, or weighted scoring functions that evaluate the current operating conditions against predetermined criteria or learning modes.

[0118] Workload information 210 can be related to... Figure 2A , Figure 2B and Figure 2C The workload information 210 described is the same as or similar to that described above. Workload information 210 may represent the characteristics or requirements of a system or application workload, which may influence communication medium selection decisions made by controller 112 or model 204 as previously described. Workload information 210 can provide context about the nature and requirements of the data traffic to be transmitted, thereby allowing the communication medium selector to align interface selection with specific application requirements or system priorities.

[0119] The workload information 210 may include one or more workload parameters 230. The workload parameters 230 may include, but are not limited to, application type identifiers (e.g., codes or tags that characterize the generating application as a video stream, database transaction, file transfer, etc.), quality of service (QoS) level indicators (e.g., priority classifications that reflect the relative importance of traffic), data rate requirements (e.g., minimum or expected bandwidth requirements expressed in bits per second), latency sensitivity metrics (e.g., maximum acceptable latency value), energy constraints (e.g., power budget allocation or battery conservation instructions), or user-defined communication policies (e.g., management rules or preferences regarding network resource utilization).

[0120] In some respects, workload parameter 230 can be derived from various sources, including application metadata, tenant service level agreements, system management instructions, or dynamic resource allocation algorithms. Workload parameter 230 can be derived from... Figure 1A The host system 156 described herein is generated and transmitted to the hybrid network interface controller via coupling interface 132, or they may be assigned by a network management system or orchestration platform outside of individual computing nodes.

[0121] exist Figure 2D The data in the data structure shown in the diagram can be used alone, or in conjunction with the selected model (such as in...). Figure 2A , Figure 2B and Figure 2CThe model 204 described herein is used in conjunction to determine how to route communication via the optical and / or wireless interfaces of the hybrid network interface controller. In some respects, parameter sets may be encoded in a structured format (e.g., JavaScript Object Notation (JSON), Type-Length-Value (TLV), Protocol Buffer (protobuf)) and exchanged between systems via control plane channels, or stored in configuration memory within the controller 112 previously described.

[0122] Now for reference Figure 3A A training system 300 is illustrated according to at least one aspect of this disclosure. The training system 300 can be configured to generate machine learning models for selecting communication media in a hybrid network interface environment. The training system 300 can operate offline or online. The training system 300 can be implemented in various network infrastructure components (including, but not limited to, controllers, orchestration platforms, and training nodes) or within a network management system.

[0123] In some aspects, training system 300 may include model trainer 302, configured to generate and refine machine learning model 304. Machine learning model 304 may be trained to predict or generate configuration information 308 based on input features. Machine learning model 304 may include various learning architectures suitable for processing network performance data and determining appropriate interface selection instructions. Machine learning model 304 may be implemented using one or more computational methods, including but not limited to supervised learning models, unsupervised clustering models, regression estimators, neural networks, decision trees, random forests, support vector machines, Bayesian networks, reinforcement learning agents, or other statistical or algorithmic predictors.

[0124] In some implementations, the machine learning model 304 can be a multi-layered neural network, where each layer includes computational nodes that transform input data through a series of mathematical operations. The neural network may include convolutional layers for feature extraction, recurrent layers for sequence processing, or transformer architectures for capturing temporal relationships in network performance data. In other implementations, the machine learning model 304 can be a decision tree or a collection of trees that partitions the input space based on feature thresholds, thereby providing interpretable decision boundaries for interface selection. In still other implementations, the machine learning model 304 may employ reinforcement learning techniques, which optimize the interface selection strategy through interaction with the network environment, learning from the results of previous selection decisions to maximize performance metrics over time.

[0125] In some aspects, the model trainer 302 may receive input data from one or more data sources 306. One or more data sources 306 may include various information repositories or data streams that provide training examples for the machine learning model 304. These data sources 306 may include real-time measurements, historical datasets, performance records, or synthetic training data, and may be collected from simulation environments, deployed devices, or external telemetry feeds. One or more data sources 306 may provide labeled examples that associate specific network conditions and workload characteristics with optimal interface selection decisions, thereby enabling supervised learning methods. In some aspects, data sources 306 may provide unlabeled data that allows the machine learning model 304 to discover patterns or clusters in network behavior, thereby supporting unsupervised or semi-supervised learning methods.

[0126] In some respects, one or more data sources 306 include or are derived from communication medium information 206 and workload information 210, which can be related to... Figures 2A-2D The communication medium information 216 and the workload data 210 described are the same or similar.

[0127] In some respects, one or more data sources 306, communication medium information 206, and workload information 210 can be provided as ground data for training the machine learning model 304. The datasets of one or more data sources 306 may include known relationships between medium conditions, workload parameters, and optimal or expected configuration decisions. Such examples can allow the machine learning model 304 to learn the appropriate mapping between inputs and outputs and to form an internal representation of how different network conditions and workload requirements should affect interface selection.

[0128] The training system 300 can utilize a loss function 310 to evaluate the performance of the machine learning model 304 during training. In some aspects, the loss function 310 compares the configuration information 308 output from the machine learning model 304 with expected or labeled configuration information (e.g., in some cases, the true value) and generates a quantitative measure of the prediction error. Depending on the nature of the prediction task and the model architecture, the loss function 310 can be implemented using various mathematical formulas. For classification tasks, such as determining which interface to activate, the loss function 310 can employ cross-entropy loss or hinge loss. For regression tasks, such as determining the percentage of bandwidth allocation, the loss function 310 can utilize mean squared error or mean absolute error. For multi-objective optimization scenarios, the loss function 310 can contain a weighted combination of multiple metrics that address competing performance objectives, such as maximizing throughput while minimizing latency or power consumption.

[0129] In some aspects, the loss function 310 generates a score or gradient that quantifies the error or discrepancy between the model prediction (e.g., configuration information 308) and the ground truth. The error signal provides feedback to guide the optimization process, allowing the model trainer 302 to adjust the internal parameters of the machine learning model 304 using techniques such as gradient descent, adaptive moment estimation, or other optimization algorithms. The model trainer 302 can use this feedback to update and improve the machine learning model 304 over multiple iterations or training cycles, thereby reducing prediction errors and improving the machine learning model's ability to make effective interface selection decisions in different network scenarios.

[0130] Once training is complete or an acceptable performance threshold is reached, the model trainer 302 produces a trained model 312. The trained model 312 represents an optimized version of the machine learning model 304, where parameters and structuring elements are configured to perform effective interface selection based on patterns learned during training. The trained model 312 may include fine-tuned weight matrices, optimized decision boundaries, or learned representations that encode relationships between network conditions, workload characteristics, and appropriate interface configuration instructions.

[0131] The trained model 312 can then be deployed to controllers, selectors, or external configuration systems (e.g., as per [reference]). Figure 2A-2C The trained model 312 (described above) is used to generate real-time configuration information for selecting the communication medium in the hybrid network interface controller. In some implementations, the trained model 312 can be compressed or quantized to reduce computational requirements while maintaining prediction accuracy, thereby enabling efficient execution on resource-constrained devices. In other implementations, the trained model 312 can be periodically updated or fine-tuned based on new data, allowing the system to adapt to evolving network conditions or workload patterns.

[0132] Now for reference Figure 3B The present disclosure describes example aspects of a trained model 312 according to at least one aspect. The trained model 312 can be used with respect to... Figure 3A The described trained model 312 is the same as or similar. Figure 3B In the example shown, the trained model 312 is configured as a neural network architecture designed to process input data 314 and generate configuration information 308 for selecting or controlling the communication medium in the hybrid network interface system.

[0133] In some aspects, input data 314 is provided to neural network 334, wherein input data 314 may include or be derived from communication medium information 206 and workload information 210, as previously stated. Figure 2A-2D and Figure 3AAs described above. Input data 314 may include various features or parameters characterizing the current state and workload requirements of the network interface. Before providing such features to the model, they can be preprocessed or transformed using techniques such as normalization, scaling, one-hot encoding, or feature extraction to improve the model's learning efficiency and generalization ability.

[0134] Input data 314 can be received by the input layer 316 of the neural network 334. The input layer 316 may include multiple input nodes 318, each representing a specific input feature or dimension. For example, each node within the input layer 316 may correspond to parameters from the communication medium information 206 or workload information 210, such as signal strength measurements, error rates, bandwidth requirements, or application type. The input layer 316 acts as the entry point for data into the neural network 334, thereby providing a structured representation of features that will influence the interface selection decision.

[0135] Each input node 318 in input layer 316 can be connected to a set of hidden layer nodes 322 via a set of weighted input-to-hidden layer connections 320. The weighted input-to-hidden layer connections 320 represent the learned association between input features and intermediate representations within the neural network 334. Each connection can be assigned a weight value that determines the strength of the signal passed from the input node to the hidden node. In some aspects, in Figure 3A The training process described herein fine-tunes weights, where the model adjusts connection strengths to minimize prediction error. Connection patterns and their associated weights encode the model's understanding of how different input features should influence node activations in subsequent layers.

[0136] The neural network 334 includes one or more hidden layers, such as layers that include hidden nodes 322 that can generate intermediate representations of the input data. The hidden nodes 322 can process their inputs, combining signals from multiple input nodes according to connection weights, and then applying nonlinear activation functions to introduce complexity into the model's computational capabilities. Common activation functions include Rectified Linear Unit (ReLU), sigmoid functions, hyperbolic tangent functions (tanh), or leaky ReLU variants, each providing different properties regarding gradient propagation and feature representation. Together, the hidden nodes 322 form a distributed representation that captures patterns, correlations, or abstractions within the input data relevant to the interface selection task.

[0137] In some respects, subsequent hidden layers, including hidden layer 326 and associated hidden nodes 324, can be included in neural network 334. Hidden nodes 324 within hidden layer 326 receive input from hidden nodes 322 of the previous layer, allowing the model to construct increasingly abstract or complex representations of the input data. By stacking multiple hidden layers, the neural network can learn hierarchical features, where earlier layers typically detect basic patterns (e.g., simple correlations between signal strength and error rate), and deeper layers combine these basic patterns into more complex concepts (e.g., complex relational patterns indicating when wireless communication may be preferred despite lower signal strength due to specific workload characteristics).

[0138] Hidden nodes 322 in the first hidden layer are connected to hidden nodes 324 in the second hidden layer 326 via a set of hidden-to-hidden-layer connections 328. The hidden-to-hidden-layer connections 328 propagate activations through the network, enabling the model to learn hierarchical or combined features related to communication behavior. Similar to the input-to-hidden-layer connections 320, weights are assigned to these connections during training, determining the strength of the influence of each node in the first hidden layer on the activation of nodes in the second hidden layer. The connection patterns between hidden layers allow the model to compose increasingly complex feature representations as information flows through the network.

[0139] The output of the final hidden layer 326 is provided to an output layer comprising one or more output nodes 332. These output nodes 332 are connected to the final hidden layer via a set of hidden-to-output layer connections 330. The output nodes 332 are configured to generate one or more outputs representing configuration information 214. This configuration information 214 may include instructions for activating a specific communication medium (e.g., optical or wireless), applying specific routing rules, or adjusting transmission behavior. Depending on the specific requirements of the interface selection task, the output layer can be constructed in various ways. For classification decisions, such as choosing between optical interface operation, wireless interface operation, or dual-interface operation, the output node 332 may generate probability distributions or discrete level assignments. For parameter tuning tasks, such as determining bandwidth allocation ratios or transmission power levels, the output node 332 may generate continuous values ​​within an appropriate range.

[0140] In some embodiments, the trained model 312 can be implemented as part of a runtime decision engine (e.g., in...). Figure 2A-2C (in the middle), or can be implemented as a trained model deployed to the control plane or network interface components (e.g., Figure 3A (312 in the model). When deployed, the trained model 312 can process the current network conditions and workload characteristics in real time, thereby generating configuration instructions based on the learned patterns and relationships to optimize interface selection.

[0141] Figure 3B The neural network architecture depicted represents one possible implementation of the machine learning model 304. Alternative architectures can be employed depending on specific requirements or constraints. For example, recurrent neural networks or long short-term memory (LSTM) networks can be used when temporal patterns are significant in network performance. Convolutional architectures can be applied when spatial or structural relationships exist in the input data. Graph neural networks can be used to model complex dependencies between network elements or communication paths. Specific architectures can be chosen based on factors such as the nature of the available data, the complexity of the decision space, computational resource constraints, or interpretability requirements.

[0142] Various regularization techniques can be applied to the trained model 312 to improve generalization performance and prevent overfitting to the training data. These techniques can include dropout (randomly deactivating nodes during training), weight decay (penalizing large weight values), batch normalization (normalizing activations between layers), or early stopping (terminating training when validation performance begins to deteriorate).

[0143] Now for reference Figure 4A An optical module 116 is depicted according to at least one aspect of this disclosure. The optical module 116 can be integrated as a component of a hybrid network interface controller described in the previous figures, such as... Figure 1A and Figure 1B The optical interface 104. The optical module 116 can provide bidirectional optical communication capability through the transmit path 402 and the receive path 404, thereby enabling the hybrid network interface controller to communicate via the optical communication medium.

[0144] In some respects, the optical module 116 can be implemented using various form factors and integration methods, including but not limited to pluggable transceiver modules, integrated photonic circuits, or discrete optoelectronic components mounted on a printed circuit board within a network interface controller. The optical module 116 can support a variety of optical transport standards, protocols, and data rates suitable for data center, enterprise, or high-performance computing environments.

[0145] In some aspects, the transmission path 402 may include a series of components configured to convert electrical signals into modulated optical signals suitable for transmission over an optical communication medium. The transmission path 402 may receive an electrical input signal 406, which may originate from MAC circuitry within a hybrid network interface controller, a controller, or a host interface. The electrical input signal 406 may represent serialized data from a higher protocol layer prepared for physical layer transmission.

[0146] An electrical input signal 406 can be provided to an encoder 408, which performs signal encoding operations to prepare electrical data for optical transmission. In some aspects, the encoder 408 can implement various encoding schemes, such as non-return-to-zero (NRZ), pulse amplitude modulation (PAM), or more advanced modulation formats. The encoder 408 can perform line encoding functions, such as 8b / 10b encoding, 64b / 66b encoding, or other data conversion techniques to ensure DC balance, clock recovery capability, and error detection. In some implementations, the encoder 408 can incorporate FEC encoding to enhance transmission reliability over optical media. The encoder 408 can be implemented using dedicated digital signal processing (DSP) circuitry, an ASIC, a field-programmable gate array (FPGA), or other suitable digital logic.

[0147] Encoded electrical signals from encoder 408 can be provided to modulator 410, which converts the electrical signals into corresponding optical signal moduli. Modulator 410 receives continuous-wave optical signals from laser 412 and modulates a carrier signal according to the encoded electrical data. In some aspects, modulator 410 can employ various modulation techniques, including but not limited to direct modulation of the laser current and external modulation using electro-optic modulators (such as Mach-Zehnder interferometers or electro-absorption modulators). The modulation scheme can be selected based on factors such as data rate, transmission distance, power consumption requirements, or integration constraints.

[0148] In some aspects, laser 412 generates a continuous-wave optical carrier signal for data transmission. Laser 412 can be implemented using various semiconductor laser technologies, including but not limited to distributed feedback (DFB) lasers, vertical-cavity surface-emitting lasers (VCSELs), or other coherent light sources suitable for fiber optic transmission. In some aspects, laser 412 can operate at specific wavelengths compatible with industry-standard optical communication windows, such as 850 nm for multimode transmission or 1310 / 1550 nm for single-mode applications. Laser 412 may include stabilization circuitry for maintaining wavelength accuracy, power control mechanisms for ensuring appropriate optical output levels, and monitoring capabilities for ensuring proper operation.

[0149] The modulated optical signal generated by the interaction between the modulator 410 and the laser 412 is directed to the optical output 414. The optical output 414 may include optical coupling elements, such as lenses, waveguides, or fiber alignment structures, which transmit the modulated light into an external optical fiber or waveguide. In some implementations, the optical output 414 may include an optical isolation component to prevent back reflections from interfering with laser operation, or a monitoring tap for sampling a portion of the output power for diagnostic purposes.

[0150] In some aspects, the receiving path 404 includes a series of components (e.g., optical processing circuitry) configured to convert incoming optical signals into electrical signals that can be processed by a hybrid network interface controller. The receiving path 404 may receive the incoming optical signal at an optical input 416, which may include optical coupling elements similar to those described for optical output 414 but configured to capture light from an external optical fiber and guide it into internal optical components of the receiving path. The incoming optical signal from optical input 416 is directed to an optical filter 418. The optical filter 418 may perform wavelength selection or noise filtering functions to improve signal quality before detection. In some aspects, the optical filter 418 may be a thin-film filter, a fiber Bragg grating, or other wavelength-selective element that allows the desired signal wavelength to pass while attenuating out-of-band noise or adjacent channel interference in a wavelength division multiplexing system. The optical filter 418 can improve receiver sensitivity by increasing the optical signal-to-noise ratio of the incoming signal.

[0151] The filtered optical signal is then directed to photodetector 420, which converts the optical signal into a corresponding current or voltage. Photodetector 420 can be implemented using various optoelectronic technologies, including but not limited to positive-intrinsic-negative (PIN) photodiodes, avalanche photodiodes (APDs), or integrated photodetectors within photonic integrated circuits. The choice of photodetector technology can depend on factors such as desired sensitivity, bandwidth, noise characteristics, or cost constraints. In some implementations, photodetector 420 may include an integrated transimpedance amplifier to convert the photogenerated current into a voltage signal.

[0152] The electrical signal generated by photodetector 420 can be provided to filter / amplifier / transimpedance amplifier (TIA) stage 422. This stage can perform multiple signal conditioning functions to prepare the detected signal for subsequent digital processing. In some aspects, filter / amplifier / TIA stage 422 may include transimpedance amplification to convert the photodetector current into a voltage with appropriate gain, equalization filtering to compensate for channel impairments or enhance signal bandwidth, automatic gain control to maintain a consistent signal level despite variations in optical input power, and / or noise filtering to improve signal quality. Filter / amplifier / TIA stage 422 can be implemented using discrete analog components, integrated analog circuitry, or mixed-signal devices that combine analog processing with digital control capabilities.

[0153] In some implementations, the filter / amplifier / TIA stage 422 signal can be further processed by an optional decoder 424 (shown in dashed lines) to indicate its optional nature in some respects. When present, the decoder 424 can perform functions such as: clock and data recovery for extracting timing information from the received signal, line decoding (e.g., 8b / 10b decoding) for reversing the encoding applied at the transmitter, and error detection or correction using forward error correction codes. The decoder 424 can be implemented using dedicated digital signal processing circuitry, an ASIC, an FPGA, or other suitable digital logic.

[0154] The processed electrical signal from the receiving path 404 can be provided as output 426. Output 426 can be provided to one or more subsequent stages of the hybrid network interface controller, such as MAC circuitry, controllers, or host interfaces, to deliver the recovered data for processing by higher-level protocols or to forward it to the host system.

[0155] Figure 4A The figure illustrates one possible implementation of an optical module architecture that provides bidirectional conversion between the electrical and optical domains. However, it should be understood that this is merely exemplary, and this disclosure considers various alternative implementations of the optical module. Such alternative implementations may include different component arrangements, additional or fewer components, alternative technologies for signal generation and detection, or variations in the signal processing chain. A specific implementation can be chosen based on factors such as cost considerations, performance requirements, size constraints, power limitations, or integration with other system components while maintaining compatibility with the overall hybrid interface architecture.

[0156] Now for reference Figure 4B Exemplary aspects of the wireless module 124B are illustrated according to at least one aspect of this disclosure. The wireless module 124B can be integrated as a component of the hybrid network interface controller described in the previous figures, such as... Figure 1A and Figure 1B The wireless interface 108. The wireless module 124B can provide bidirectional wireless communication capability through the transmit path 428 and the receive path 430, enabling the hybrid network interface controller to communicate via the wireless communication medium.

[0157] In some respects, the wireless module 124B can be implemented using various form factors and integration methods, including but not limited to radio frequency integrated circuits (RFICs), system-on-a-chip (SoCs) with integrated RF components, or discrete RF components mounted on a printed circuit board within a network interface controller. The wireless module 124B can support a variety of wireless transmission standards, protocols, and data rates suitable for data center, enterprise, or high-performance computing environments.

[0158] Transmit path 428 includes a series of components configured to generate and transmit RF signals based on electrical input data. Transmit path 428 receives electrical input signals as input 432, which may originate from MAC circuitry within the hybrid network interface controller, the controller, or the host interface. The electrical input signals may represent serial data from a higher protocol layer prepared for physical layer transmission.

[0159] An electrical input (e.g., input 432) can be provided to an encoder 434, which performs signal encoding operations to prepare electrical data for wireless transmission. In some aspects, the encoder 434 can implement various encoding schemes, such as binary phase shift keying (BPSK), quadrature phase shift keying (QPSK), quadrature amplitude modulation (QAM), or more advanced modulation formats such as orthogonal frequency division multiplexing (OFDM). The encoder 434 can perform functions such as FEC, interleaving, framing, or other data conversion techniques to enhance transmission reliability over wireless media. In some implementations, the encoder 434 can incorporate protocol-specific framing for standards such as IEEE 802.11 (Wi-Fi), 5G NR, or proprietary protocols. The encoder 434 can be implemented using dedicated DSP circuitry, an ASIC, an FPGA, or other suitable digital logic.

[0160] The encoded signal from encoder 434 can then be delivered to RF up-converter 436. RF up-converter 436 can convert the baseband or intermediate frequency (IF) signal to a higher carrier frequency suitable for wireless transmission. In some aspects, RF up-converter 436 may include one or more mixer stages, filter elements, and impedance matching networks to ensure correct signal conversion to the desired RF band. Depending on the specific implementation and application requirements, RF up-converter 436 can operate in a variety of frequency bands, including but not limited to sub-6 GHz, millimeter wave (e.g., 24-100 GHz), or terahertz frequencies.

[0161] The RF upconverter 436 can receive a frequency reference from a signal source 438. The signal source 438 may include a local oscillator, a phase-locked loop (PLL), a clock generator, and / or a frequency synthesizer that provides a carrier frequency for the upconversion operation. In some aspects, the signal source 438 may be tunable to support multiple frequency channels or bands. The signal source 438 may include frequency stabilization circuitry for maintaining accurate carrier frequency generation, phase noise reduction components for improving signal quality, and calibration mechanisms for ensuring compliance with regulatory requirements or protocol specifications.

[0162] The up-converted RF signal from RF up-converter 436 is provided to amplifier 440. Amplifier 440 increases the power level of the RF signal to ensure sufficient transmission strength. In some aspects, amplifier 440 may include one or more amplification stages, such as driver amplifiers, followed by power amplifiers (PAs), to achieve the required transmission power while maintaining signal linearity and efficiency. Amplifier 440 may incorporate power control mechanisms to adjust the output level based on transmission distance, link quality requirements, or power-saving instructions from the controller. In some implementations, amplifier 440 may include linearization techniques, such as digital predistortion or envelope tracking, to improve efficiency at higher power levels.

[0163] The amplified RF signal from amplifier 440 is then directed to wireless output 442. Wireless output 442 can also be provided to impedance matching networks, filters, and / or transmission line structures that transfer RF energy to external antennas (e.g., antenna 129) or transmission structures. In some aspects, wireless output 442 can be provided to one or more diagnostic capabilities to monitor forward and reflected power levels for link quality assessment or protection functions.

[0164] In some aspects, the receive path 430 includes components (e.g., wireless processing circuitry) configured to receive incoming RF signals and convert them into baseband digital data suitable for processing by a hybrid network interface controller. The receive path 430 can accept incoming wireless signals as input 444. Input 444 can be routed to / from one or more RF coupling elements, impedance matching structures, or filtering components (configured to effectively capture RF signals from external antennas and direct them into internal components of the receive path).

[0165] The incoming RF signal received from input 444 is then directed to LNA 450. LNA 450 amplifies the weak received RF signal while introducing minimal additional noise. LNA 450 can incorporate automatic gain control mechanisms to adapt to varying input signal strengths while preventing saturation in subsequent stages. In some implementations, LNA 450 may include multiple gain stages with one or more filtering stages to provide amplification and selectivity.

[0166] The amplified RF signal from the LNA 450 can be directed to the RF down-converter 446. The RF down-converter 446 can convert the received signal from its RF carrier frequency to an intermediate frequency (IF) or directly to baseband for subsequent processing. In some aspects, the RF down-converter 446 may include a mixer stage, filter elements, and gain control mechanisms to ensure proper signal conversion while maintaining signal quality. The RF down-converter 446 can utilize the same or a separate frequency reference as the transmission path to ensure coherent demodulation.

[0167] The down-converted signal from RF down-converter 446 can be provided to amplifier 448. Amplifier 448 increases the signal level of the down-converted signal to ensure detection and processing. In some aspects, amplifier 448 may include a low-noise amplifier (LNA) for improving sensitivity while minimizing noise contribution, automatic gain control (AGC) circuitry for adapting to different input signal strengths, or a variable gain stage for optimizing the signal level for subsequent processing. Amplifier 448 can be implemented using various RF amplification techniques selected based on noise figure, linearity, power consumption, or integration requirements.

[0168] The amplified signal from amplifier 448 is then directed to demodulator 452. Demodulator 452 can extract the baseband data signal from the modulated waveform according to the modulation scheme used by the transmitter. In some respects, demodulator 452 can perform phase recovery, symbol timing recovery, equalization, or other signal processing functions to compensate for channel impairments and improve detection accuracy. Depending on the specific requirements and constraints of wireless module 124B, demodulator 452 can be implemented using analog circuitry, digital signal processing techniques, or a combination of both.

[0169] The demodulated signal from demodulator 452 is provided to decoder 454. Decoder 454 processes the demodulated signal to recover the digital data stream originally encoded at the transmitter. In some respects, decoder 454 may perform error correction using forward error correction codes, perform deinterleaving to reverse the interleaving applied at the transmitter, frame alignment to identify packet boundaries, or perform protocol-specific processing to extract payload data. Decoder 454 may be implemented using dedicated digital signal processing circuitry, ASIC, FPGA, or other suitable digital logic.

[0170] The output of decoder 454 is provided as output 456. Output 456 can be provided to one or more subsequent stages of the hybrid network interface controller, such as MAC circuitry, controllers, or host interfaces, to deliver the recovered data for processing by higher-level protocols or to forward it to the host system.

[0171] The wireless module 124B may support one or more wireless communication standards, including but not limited to IEEE 802.11 (Wi-Fi), 5G NR, millimeter-wave wireless, or proprietary high-frequency protocols suitable for data center or enterprise applications. In some implementations, the wireless module 124B may be reconfigurable to adapt to different protocols, frequency bands, or operating modes based on controller instructions or system requirements.

[0172] Figure 4BThe figure illustrates one possible implementation of a wireless module architecture providing bidirectional wireless communication capabilities. However, it should be understood that this is merely exemplary, and this disclosure considers various alternative implementations of the wireless module. Such alternative implementations may include different component arrangements, additional or fewer components, alternative radio frequency technologies, different frequency bands, or variations in the signal processing chain. A specific implementation can be selected based on factors such as performance targets, integration methods, cost considerations, or power efficiency requirements, while maintaining compatibility with the overall hybrid interface architecture.

[0173] Now for reference Figure 4C Another wireless module 124C is described according to at least one aspect of this disclosure. Figure 4C The wireless module 124C depicted represents an example implementation configured to generate and process high-frequency wireless signals using photonics technology. Photonics-based methods can be integrated as components of the hybrid network interface controller described in the previous figures, such as... Figure 1A and Figure 1B The wireless interface 108. The wireless module 124C provides bidirectional wireless communication capability through wireless transmission path 458 and wireless reception path 460 (each employing optoelectronic components to perform high-frequency signal generation and down-conversion).

[0174] The wireless module 124C can be implemented using various integration methods, including but not limited to integrated photonic circuits, hybrid photonic-electronic components, silicon photonic platforms, or discrete optoelectronic components mounted on a printed circuit board within the network interface controller. In some aspects, Figure 4C The photonic mixing architecture described herein can be beneficial for systems requiring wideband operation and / or high-frequency carrier generation with low phase noise. The wireless module 124C can support a variety of wireless transmission standards and protocols suitable for high-speed data transmission in data center, enterprise, or high-performance computing environments, particularly in millimeter-wave and terahertz bands where conventional electronic methods may face technical limitations.

[0175] In the wireless transmission path 458, an electrical input signal can be provided as input 432, which can be associated with digital data from the MAC circuitry within the hybrid network interface controller, the controller, or the host interface. Input 432 can be directed to appropriate impedance matching, level shifting, or buffering circuitry to ensure proper signal integrity when transmitting data from the digital processing circuitry to the photonic transmission path components.

[0176] An electrical input signal (e.g., input 432) can be provided to an encoder 434, which performs signal encoding operations to prepare electrical data for wireless transmission. The encoder 434 can implement various encoding schemes, such as BPSK, QPSK, QAM, or other modulation formats suitable for high-speed wireless transmission. In some aspects, the encoder 434 can perform functions such as FEC, interleaving, framing, or other data conversion techniques to enhance the reliability of transmission over a wireless medium. The encoder 434 can be implemented using dedicated DSP circuitry, an ASIC, an FPGA, or other suitable digital logic.

[0177] The encoded signal from encoder 434 is provided to modulator 462, which modulates optical carrier 464 based on the encoded data. Modulator 462 may include an electro-optic modulator, such as a Mach-Zehnder modulator, an electroabsorption modulator, or other devices capable of applying encoded data to the optical carrier. In some aspects, modulator 462 may employ direct modulation techniques, where the drive current of the laser is directly modulated according to the data signal; or external modulation, where an external device modulates the continuous-wave optical signal. Modulator 462 may support various modulation formats, including intensity modulation, phase modulation, or composite modulation schemes that encode multiple bits per symbol to improve spectral efficiency.

[0178] The optical carrier 464 provides a continuous-wave laser signal suitable for photonic upconversion. The optical carrier 464 can be generated by a laser source, such as a DFB laser, an external cavity laser, or a tunable laser that produces a spectrally pure optical tone. In some implementations, the optical carrier 464 can operate at a specific wavelength (e.g., 1310 nm or 1550 nm) compatible with a photonic integrated circuit platform, providing low propagation loss in the optical waveguide and fiber. The optical carrier 464 may include stabilization circuitry for maintaining wavelength accuracy, power control mechanisms for ensuring appropriate optical output levels, and monitoring capabilities for ensuring proper operation.

[0179] The modulated optical signal from modulator 462 can be directed to photonic mixer stage 465, which may include UTC photodiode 466. UTC photodiode 466 may be a high-speed photodetector optimized for high-frequency applications and includes a structure for electrons to pass through the depletion region, resulting in higher bandwidth capability compared to conventional PIN photodiodes. UTC photodiode 466 performs photonic mixing of the modulated carrier and optical reference signal 468 to generate a high-frequency electrical RF signal. This photonic mixing process utilizes optical heterodyne, where the beat frequencies of two optical frequencies with a controlled frequency difference generate an electrical signal at their difference frequency. Such signals can be generated in the millimeter-wave or terahertz range.

[0180] Optical reference signal 468 can provide a second optical tone, which, when mixed with the modulated optical carrier in UTC photodiode 466, generates the desired RF carrier frequency. Optical reference signal 468 can be generated by a separate laser source, derived from an optical frequency comb generator, or created using an optical synthesizer that provides frequency control. The frequency difference between optical carrier 464 and optical reference signal 468 determines the carrier frequency of the generated RF signal. In some implementations, one or both optical signals can be wavelength tunable to allow for adjustable carrier frequency generation, thus providing flexibility in wireless channel selection or multi-band operation.

[0181] The RF signal generated by the UTC photodiode 466 is provided to amplifier 470, which amplifies the signal power to a level suitable for wireless transmission. Amplifier 470 may include one or more RF amplification stages, such as a preamplifier followed by a power amplifier, to achieve the required transmission power while maintaining signal linearity and efficiency. In some aspects, amplifier 470 may incorporate technologies such as gallium nitride (GaN) or silicon-germanium (SiGe) to provide efficient amplification at millimeter-wave or Asia-Pacific Hertz frequencies. Amplifier 470 may include a power control mechanism to adjust the output level based on transmission distance requirements, link quality targets, or power-saving instructions from the controller.

[0182] The amplified RF signal from amplifier 470 is provided as wireless output 472. Wireless output 472 can be provided to one or more impedance matching networks, filters, and / or transmission line structures that efficiently transfer RF energy to external antennas (e.g., antenna 129), waveguides, or other radiating structures. In some aspects, wireless output 472 can be provided to one or more coupling elements optimized for millimeter-wave or terahertz frequencies, such as waveguide transitions, probe feeds, or dedicated RF connectors that maintain signal integrity at high frequencies. Wireless output 472 can be combined with diagnostic capabilities to monitor forward and reflected power levels for link quality assessment or protection functions.

[0183] The wireless receiving path 460 is configured to perform photonic down-conversion on the received RF signal using electro-optic modulation technology. The received RF signal is received as input 444, which can be routed to / from RF coupling elements, impedance matching structures, or filtering components within the internal components of the receiving path, configured to efficiently capture RF signals from an external antenna and direct them. Input 444 is routed to / from one or more pre-filters to reject out-of-band interference or image frequencies that may degrade receiver performance.

[0184] The RF signal received from input 444 can be provided to LNA 474. LNA 474 amplifies the weak received RF signal while introducing minimal additional noise. In some implementations, LNA 474 can be implemented using semiconductor technologies optimized for low-noise performance at high frequencies, such as indium phosphide (InP), gallium arsenide (GaAs), or SiGe processes. LNA 474 can incorporate automatic gain control mechanisms to adapt to varying input signal strengths while preventing saturation of subsequent stages. In some implementations, LNA 474 may include multiple gain stages with one or more filtering stages to provide amplification and selectivity.

[0185] An amplified RF signal from LNA 474 is applied to electro-optic modulator 478, which is also supplied with an optical signal 476. Electro-optic modulator 478 converts the RF signal into optical domain modulation, thereby effectively encoding RF information onto an optical carrier. This technique utilizes the high bandwidth capability of optical elements to process high-frequency RF. Depending on bandwidth, insertion loss, or integration requirements, electro-optic modulator 478 can be implemented using various techniques, such as lithium niobate modulators, silicon photonic modulators, or polymer-based electro-optic devices.

[0186] Optical signal 476 provides a continuous-wave laser signal that serves as a carrier wave for the electro-optic modulation process. Optical signal 476 can be generated from a laser source similar to those described for optical carrier 464 (selected for spectral purity, stability, and compatibility with electro-optic modulator 478). In some implementations, optical signal 476 can be derived from the same laser source used in the transmission path to enhance system coherence or reduce component count.

[0187] The modulated optical signal from the electro-optic modulator 478 is directed to the photodetector 480, which converts the optical modulation back into an electrical signal. The photodetector 480 can be implemented using various optoelectronic technologies, such as PIN photodiodes, avalanche photodiodes, or dedicated high-speed photodetectors optimized for the chosen modulation format and frequency range. The photodetector 480 effectively performs the photon down-conversion process, generating an electrical signal that contains information from the original RF wireless signal but converted to a lower frequency range more suitable for processing by conventional electronic devices.

[0188] The electrical output from photodetector 480 is provided to filter / amplifier stage 482. Filter / amplifier stage 482 performs signal conditioning functions to prepare the detected signal for subsequent processing. In some aspects, filter / amplifier stage 482 may include transimpedance amplification for converting the photodetector current into a voltage with appropriate gain, equalization filtering for compensating for channel impairments or enhancing signal bandwidth, automatic gain control for maintaining a consistent signal level, and noise filtering for improving signal quality. Filter / amplifier stage 482 may be implemented using discrete analog components, integrated analog circuitry, or mixed-signal devices that combine analog processing with digital control capabilities.

[0189] The regulated signal from filter / amplifier stage 482 is provided to demodulator 484. Demodulator 484 extracts the baseband or intermediate frequency (IF) signal from the received waveform according to the modulation scheme employed by the transmitter. In some respects, demodulator 484 may perform phase recovery, symbol timing recovery, equalization, or other signal processing functions to compensate for channel impairments and improve detection accuracy. Depending on the specific requirements and constraints of wireless module 124, demodulator 484 may be implemented using analog circuitry, digital signal processing techniques, or a combination of both.

[0190] The demodulated signal from demodulator 484 is provided to decoder 486. Decoder 486 processes the demodulated signal to recover the digital data stream originally encoded at the transmitter. In some respects, decoder 486 may perform error correction using forward error correction codes, perform deinterleaving to reverse the interleaving applied at the transmitter, perform frame alignment to identify packet boundaries, or perform protocol-specific processing to extract payload data. Decoder 486 may be implemented using dedicated digital signal processing circuitry, ASIC, FPGA, or other suitable digital logic.

[0191] The decoded digital signal from decoder 486 is provided as output 488. Output 488 can be provided to one or more subsequent stages in the hybrid network interface controller, such as MAC circuitry, controllers, or host interfaces, thereby delivering the recovered data for higher-level protocol processing or forwarding to the host system. Figure 4C The photonics-based wireless module 124 depicted in the diagram enables ultra-wideband and high-frequency wireless communication, offering potential advantages in spectral purity, frequency stability, and bandwidth compared to conventional all-electronic methods. The optical carrier and modulator provide high-frequency tunability and are integrated with other optical components within the hybrid network interface controller.

[0192] Figure 4C The illustrated photonics-based wireless module 124 represents one possible implementation of ultra-wideband and high-frequency wireless communication. However, it should be understood that this is merely exemplary, and this disclosure contemplates various alternative implementations of the photonics-based wireless module. Such alternative implementations may include different component arrangements, additional or fewer components, alternative photonic mixing techniques, different modulation schemes, or variations in photoelectric conversion methods. A specific implementation can be selected based on factors such as frequency requirements, bandwidth demands, power constraints, or integration considerations, while maintaining the ability to utilize the advantages of photonic signal generation for wireless communication within a hybrid network interface architecture.

[0193] Now for reference Figure 4D A hybrid optical transmission system 490 is described according to at least one aspect of this disclosure. The hybrid optical transmission system 490 can be configured to generate and selectively route modulated optical signals for use in conventional optical communication or photonic RF signal generation. The hybrid optical transmission system 490 can be described with reference to previous figures (e.g., ...). Figure 1A-1B The hybrid network interface controller described herein may be implemented within the hybrid network interface controller, or as part of the previously described optical module 116 or wireless module 124. In some aspects, the hybrid optical transport system 490 provides a unified signal generation architecture that supports both optical and wireless transport paths, thereby enhancing system flexibility and resource utilization.

[0194] The hybrid optical transmission system 490 receives an electrical input signal as input 432, which can be compared with a reference signal. Figure 4B and Figure 4C The input 432 described herein is the same as or similar to that described herein. Input 432 may represent digital data from MAC circuitry within a hybrid network interface controller, a controller, or a host interface. In some aspects, input 432 may be provided to one or more suitable signal conditioning circuitry, such as impedance matching networks, level shifters, or input buffers, to ensure proper signal integrity when transmitting data from digital processing circuitry to optical modulation components. Depending on specific implementation requirements, input 432 may have various input formats, including serial data streams, parallel data buses, or differential signaling. In some aspects, input 432 may include protocol-specific framing or formatting logic according to the relevant communication standard prior to encoding.

[0195] Input 432 is provided to encoder 434, which can communicate with about Figure 4B and Figure 4C The encoder 434 described is identical or similar. Encoder 434 performs signal encoding operations to prepare electrical data for optical modulation. In some aspects, encoder 434 can implement various signal processing operations, including but not limited to data framing, scrambling, encryption, FEC, interleaving, or protocol-specific encoding. Encoder 434 can generate an encoded data stream suitable for optical modulation, possibly in conjunction with line coding schemes such as 8b / 10b, 64b / 66b, or more advanced modulation formats, depending on system requirements and target performance metrics. In some implementations, encoder 434 can support multiple encoding modes or adaptive encoding parameters that can be dynamically configured based on operating conditions, quality of service requirements, or system instructions. Encoder 434 can be implemented using dedicated DSP circuitry, ASIC, FPGA, or other suitable digital logic.

[0196] The encoded signal from encoder 434 is provided to modulator 462, which can be used with respect to... Figure 4C The modulator 462 described is the same as or similar to the one described above. Modulator 462 can be configured to convert encoded electrical data into corresponding optical signal variations. Modulator 462 receives an optical carrier 464, which can be used with respect to… Figure 4CThe optical carrier 464 described is the same as or similar to the optical carrier 464 described above. The optical carrier 464 may include a continuous-wave laser signal generated by a laser diode, semiconductor laser, or other coherent light source. In some implementations, the optical carrier 464 may be generated by a light synthesizer, wavelength-tunable laser, or frequency-locked light source that provides accurate spectral characteristics suitable for both optical communication and photonic mixing applications. The optical carrier 464 may operate at wavelengths compatible with optical communication standards, such as 850 nm for multimode transmission or 1310 / 1550 nm for single-mode applications. In some aspects, the optical carrier 464 may include stabilization circuitry for maintaining wavelength accuracy, power control mechanisms for ensuring appropriate optical output levels, and monitoring capabilities for ensuring proper operation.

[0197] Modulator 462 modulates optical carrier 464 based on the encoded signal, thereby generating a modulated optical signal carrying the encoded information. For example... Figure 4D As illustrated, the photonic signal generator may include an encoder 434, an optical modulator 462, and an optical carrier 464. The modulated optical output from the modulator 462 is directed to a branch module 492. The branch module 492 allows for the selective routing of the modulated optical signal to different system outputs according to operational requirements. The branch module 492 can be implemented using various optical components, including but not limited to optical switches, splitters, directional couplers, wavelength-selective elements, or photonic integrated circuit routing structures. In some aspects, the branch module 492 may incorporate active components, such as electro-optic switches or thermo-optical tuning elements that allow dynamic control of optical path selection. Alternatively, the branch module 492 may utilize passive optical splitters with downstream optical path selection or attenuation to guide optical power, depending on system requirements. The branch module 492 may be controlled by digital signals from a system controller, configuration memory, or other control logic that determines the appropriate routing path based on the current operating mode, performance requirements, or fault conditions.

[0198] Branch module 492 can be configured to route the modulated optical signal along one of a plurality of selectable paths, thereby enabling hybrid optical transport system 490 to support dual functionality through shared front-end processing. In the first path, the modulated optical signal can be directed to optical output 414, which can be connected to... Figure 4AThe described optical output 414 is identical or similar. Optical output 414 is configured to transmit standard light to a remote receiver via a fiber optic connector. When a modulated optical signal is routed to optical output 414, the hybrid optical transmission system 490 operates as a conventional optical transmitter, enabling communication via optical fiber or other waveguide media. Optical output 414 may include coupling elements, such as lenses, waveguides, or fiber alignment structures, which efficiently transmit the modulated light into an external optical fiber or waveguide. In some implementations, optical output 414 may include optical isolation components to prevent back reflections from interfering with laser operation.

[0199] In an alternative path, the modulated optical signal can be delivered to a single-line carrier (UTC) photodiode 466, which can be coupled with... Figure 4C The UTC photodiode 466 described is identical or similar to the UTC photodiode 466. The UTC photodiode 466 is configured for photonic RF generation, converting optical signals into high-frequency electrical signals suitable for wireless transmission or further RF processing. When the modulated optical signal is routed to the UTC photodiode 466, the hybrid optical transmission system 490 is a photonic wireless signal generator, enabling communication via millimeter-wave or terahertz wireless channels. The UTC photodiode 466 modulates the carrier wave with a reference optical signal (…). Figure 4D Not shown in the text, but about Figure 4C (As described) Photonic mixing is performed to generate a high-frequency electrical RF signal. The photonic mixing process can utilize optical heterodyne, where the beat frequencies of two optical frequencies with a controlled frequency difference generate an electrical signal at their difference frequency.

[0200] UTC photodiode 466 and light output 414 in Figure 4D The dashed lines used in the diagrams indicate that their use can be optional or dynamically selected based on system configuration, where system components can be activated or deactivated according to operational requirements, thereby saving power or resources when certain functions are not needed.

[0201] Figure 4E yes Figure 4C and / or Figure 4D A simplified schematic diagram of the photonic mixer stage 465 mentioned above. In some aspects, the photonic mixer stage 465 includes a UTC photodiode 466 that receives two optical input signals, labeled optical signal 1 and optical signal 2. These signals can originate from separate continuous-wave laser sources, and their optical frequency difference defines the output signal frequency. The UTC photodiode 466 performs optical heterodyne operation to generate an electrical signal with a frequency corresponding to the beat frequency between the two optical inputs. This high-frequency electrical signal can be used to generate wireless transmissions in the millimeter-wave or terahertz bands in a hybrid network interface controller.

[0202] Now for reference Figure 5AAccording to at least one aspect of this disclosure, a flowchart of an example method 500 for dynamically selecting or modifying an active communication interface based on monitored communication and workload conditions is depicted. Method 500 can be provided by a hybrid network interface controller or a controller within a device (such as reference ... Figure 1A-2D The controller 112 described herein is used to achieve adaptive selection between the optical communication interface and the wireless communication interface.

[0203] Method 500 begins in a start box 502, which represents the initialization of the interface selection process. The start box 502 can be triggered by various events or in response to specific application requests, including but not limited to: system startup, periodic evaluation intervals, detection of changed network conditions, and receiving configuration commands from the host system or orchestrator. In some implementations, the start box 502 may include initializing monitoring components, loading threshold parameters from configuration memory, or resetting state variables used in subsequent decision-making processes.

[0204] From the start box 502, method 500 proceeds to box 504, where the system receives communication performance parameters and workload information. Box 504 represents the data collection phase, during which the controller collects relevant metrics and contextual information necessary to make informed interface selection decisions. Communication performance parameters may include a set of different measurements and metrics characterizing the current operational state of the optical and wireless interfaces.

[0205] As previously described, communication performance parameters may include, but are not limited to: SNR measurements quantifying the ratio of desired signal power to background noise power; delay measurements indicating transmission delay on the communication medium; throughput measures representing the actual data transmission rate achieved; congestion indicators showing resource contention or bandwidth saturation; and error rate values ​​(such as BER or packet error rate (PER)) for optical and wireless communication links. In some implementations, communication performance parameters may include statistical aggregations that provide historical context beyond instantaneous measurements, such as moving averages, variance metrics, or trend analyses.

[0206] As previously described, the workload information collected in box 504 may include parameters characterizing the requirements, priorities, or constraints of the data streams that will traverse the communication interface. This workload information may indicate attributes associated with a given data stream, application, or virtual machine, such as the data's priority level (e.g., mission-critical versus background transmission), latency sensitivity (e.g., real-time applications versus bulk data transmission), or bandwidth requirements (e.g., high-definition video streaming versus telemetry data). The workload information may be obtained from various sources, including packet header fields, Quality of Service (QoS) labels, application profiles, or explicit instructions from a workload management system. In some aspects, the workload information may include forward-looking metrics that allow for proactive interface selection before actual data transmission begins, such as anticipated traffic surges, pre-scheduled data transmissions, or resource reservations.

[0207] The collection mechanism implemented in box 504 may include direct polling of interface hardware registers, receiving telemetry data from a monitoring agent, sampling traffic statistics from a buffer management system, or processing control plane messages from an external orchestration system. In some implementations, data collection may be performed at different sampling rates for different parameters, with some metrics being captured at a higher frequency than others that may change more slowly.

[0208] After collecting communication performance parameters and workload information in box 504, method 500 proceeds to box 506, where the system compares the communication performance parameters and workload information against one or more defined thresholds. Box 506 represents the evaluation phase, during which the collected data is analyzed against established criteria to determine appropriate interface configuration actions.

[0209] The thresholds mentioned in Box 506 can include various types of decision criteria. These thresholds can be statically configured through management policies or device configurations, dynamically calculated based on the current system state or resource availability, or learned through machine learning processes that analyze historical behavior to identify optimal operating points. Thresholds can include simple numerical cutoffs (e.g., "Activate the radio interface if the optical link BER exceeds 10^-9"), conditional expressions combining multiple parameters (e.g., "Switch to optical-only mode if the radio SNR falls below 15 dB AND (and) the optical link utilization is below 60%), or complex decision boundaries generated by machine learning models, such as those concerning... Figure 3A and Figure 3B As mentioned above.

[0210] In some implementations, the comparison operation in box 506 may involve multiple threshold evaluations performed in parallel, where a weighted scoring system or decision tree determines the final result based on a combination of individual comparisons. When measurements fluctuate near threshold boundaries, the comparisons may incorporate hysteresis mechanisms to prevent rapid oscillations between interfaces, or include a damping factor that considers the duration of conditional stability before triggering configuration changes.

[0211] Following the threshold comparison in block 506, method 500 reaches decision points 508 and 510, which represent branch points in the control flow based on the evaluation results. At decision point 508, the system determines whether to activate or modify the wireless interface based on the comparison results from block 506. Similarly, at decision point 510, the system determines whether to activate or modify the optical interface based on the comparison results.

[0212] Decision points 508 and 510 can be implemented as independent evaluations rather than mutually exclusive branches, which can lead to concurrent actions on both interfaces. This independence enables the system to support various operating modes, including single-interface operation (optical or wireless), dual-interface operation with traffic separation, or load balancing across both interfaces.

[0213] The activation or modification actions determined at decision points 508 and 510 may include enabling or disabling a specific interface (e.g., powering on or off interface circuitry), adjusting interface operating parameters (e.g., modifying transmit power levels, changing modulation schemes, updating protocol timing parameters), or switching between operating modes (e.g., switching from single-interface to dual-interface operation, or vice versa). In some implementations, the actions may include finer-grained configurations, such as activating a specific channel or wavelength within an optical interface, or selecting a specific frequency band or spatial stream within a wireless interface.

[0214] After making interface configuration decisions at points 508 and 510, method 500 proceeds to box 512, where traffic is routed through the network interface controller based on the activated or modified interface configuration. Box 512 represents the implementation phase of establishing data flow paths based on the selected interface configuration. Routing operations may involve configuring packet bootstrapping rules, updating forwarding tables, programming switching elements, or adjusting buffer allocations to guide traffic flows through the appropriate interface.

[0215] In some implementations, traffic routing in box 512 may include flow-based differentiation, where different data flows are routed to different interfaces based on their characteristics or requirements. For example, latency-sensitive traffic may be routed to the interface that provides the lowest latency, while bulk data transfers may be routed to the interface with the highest available bandwidth. The routing configuration may incorporate a load balancing algorithm that distributes traffic across multiple active interfaces to optimize resource utilization, or a redundancy mechanism that replicates critical traffic across two interfaces to improve reliability.

[0216] Method 500 ends in box 514, which signifies the completion of the current interface selection and configuration cycle. In some implementations, reaching box 514 may trigger a delay interval before restarting the method in box 502, thereby establishing a periodic evaluation pattern. Alternatively, the system may remain in box 514 until an event-based trigger initiates a new evaluation cycle, such as detecting changed network conditions, receiving new workload information, or an explicit reconfiguration command.

[0217] It should be noted that method 500 can operate autonomously within the network interface controller as a self-contained control loop, or it can be notified by higher-level control signals from the host system, network orchestrator, or management plane. Interface selection decisions can incorporate broader contextual information beyond immediate performance parameters and workload characteristics, including system-level considerations such as energy efficiency targets, thermal management constraints, or fault tolerance requirements.

[0218] In some respects, dual-mode operation (e.g., when both the optical and wireless interfaces are active simultaneously) can be triggered based on considerations beyond simple threshold evaluation. Such considerations might include resource availability across the broader network, redundancy requirements of mission-critical applications, or policy-driven standards established through management configuration. Dual-mode operation can be implemented in a variety of configurations, including an active-active setup where both interfaces concurrently carry traffic, an active-standby setup where one interface serves as a backup, or a dedicated setup where the interface is devoted to a specific traffic category.

[0219] Now for reference Figure 5B The present disclosure describes a flowchart of an example method 516 for generating communication interface configuration output using a decision model, according to at least one aspect of the present disclosure. Method 516 can be provided by a hybrid network interface controller or a controller within a device (such as a reference cipher). Figure 1A-2D The controller 112 described herein, or as per reference, Figure 2C The external configuration generation system described is used for implementation. Method 516 can achieve model-driven selection between optical communication interfaces and wireless communication interfaces based on workload characteristics and communication medium performance metrics.

[0220] Method 516 begins in a start box 518, which represents the initialization of the model-based configuration generation process. The start box 518 can be triggered by various events, including but not limited to: system startup, periodic evaluation intervals, detection of changed network conditions, receipt of new workload information, or explicit configuration update requests. In some implementations, the start box 518 may include initializing model parameters, loading trained model weights from non-volatile memory, or preparing input feature vectors for subsequent processing operations.

[0221] From the start box 518, method 516 proceeds to box 520, where the system receives workload information. Box 520 represents the data acquisition phase, during which the system collects contextual information related to the data traffic that will traverse the communication interface. The workload information collected at box 520 may include parameters characterizing the requirements, priorities, or constraints of the current or anticipated data flow.

[0222] In some aspects, workload information may include priority level metrics that categorize data according to operational importance (e.g., mission-critical, business-critical, or back-end), latency sensitivity measures that quantify time constraints for different applications (e.g., real-time, interactive, or batch processing), bandwidth requirement estimates that predict data transfer volume requirements (e.g., peak rate, sustained throughput, or surge patterns), or application-specific requirements (such as security, reliability, or energy efficiency constraints). Workload information may also include traffic type classifications (e.g., streaming media, database transactions, file transfers, or control messages), session duration predictions indicating expected connection lifetimes, or Quality of Service (QoS) parameters that establish performance expectations for different data categories.

[0223] Workload information can be obtained from various sources, including but not limited to: packet inspection mechanisms that analyze header fields or traffic patterns; application programming interfaces (APIs) that receive explicit statements from software components; workload management systems that track resource allocation and reservation; virtual machine managers or container orchestrators that monitor application behavior; or policy engines that apply management commands to different traffic levels. In some implementations, workload information may include historical patterns or statistical models that predict future resource demands based on observed behavior, enabling proactive configuration adjustments before actual demand materializes.

[0224] After acquiring workload information in box 520, method 516 proceeds to box 522, where the system receives communication medium information. Box 522 represents a supplementary data collection phase focusing on the operational status and performance characteristics of available communication interfaces. The communication medium information collected in box 522 may include measurements and indicators characterizing the current operational status of the optical and / or wireless interfaces.

[0225] In some aspects, communication medium information may include metrics such as: signal quality measurements that quantify transmission fidelity (e.g., signal-to-noise ratio, bit error rate, or eye diagram parameters), bandwidth availability indicators that assess current capacity (e.g., link utilization percentage, available data rate, or congestion indicator), delay measurements that characterize transmission delay (e.g., round-trip time, jitter, or propagation delay), or reliability statistics that reflect link stability (e.g., packet loss rate, retransmission count, or link state transitions). Communication medium information may also include power consumption data associated with each interface, spectral efficiency measures of wireless communication, attenuation measurements of optical links, or environmental factors affecting performance, such as temperature, vibration, or electromagnetic interference levels.

[0226] Communication medium information can be obtained through various monitoring mechanisms, including but not limited to: hardware registers that expose physical layer statistics, dedicated telemetry components that measure signal characteristics, performance counters that track protocol-level events, or diagnostic routines that actively probe interface capabilities. In some implementations, the communication medium information may include predictive elements derived from trend analysis, seasonal patterns, or environmental correlations, which anticipate potential performance changes before they manifest as observable degradation.

[0227] After collecting workload and communication medium information in blocks 520 and 522, method 516 proceeds to block 524, where the system inputs the collected features into the model. In block 524, contextual information about workload requirements and interface capabilities is processed through a structured decision framework to determine the optimal interface selection.

[0228] The model mentioned in box 524 can be a model configured to transform input features into configuration decisions based on learning patterns, programming logic, or optimization strategies. In some aspects, the model can be implemented as a rule-based decision engine that applies conditional logic to input parameters, a decision tree that recursively partitions the feature space according to value thresholds, a statistical classifier that assigns configuration results based on feature distributions, or a machine learning model (such as a neural network, support vector machine, or ensemble method) that utilizes supervised learning to capture the complex relationship between inputs and results. The model architecture can be selected based on factors such as interpretability requirements, computational efficiency constraints, or accuracy goals for different operational scenarios.

[0229] In some implementations, the model may have used historical data that correlates workload characteristics and communication medium states with optimal interface configurations, similar to a reference model. Figure 3A The aforementioned processes involve training. Such training can employ supervised learning techniques (where ground truth labels are obtained from self-observed performance results), reinforcement learning methods that optimize configuration strategies through iterative interaction with the environment, or hybrid strategies that combine domain expertise with data-driven discovery. The model can incorporate various regularization techniques to improve generalization performance under different operating conditions, prevent overfitting to training examples, or adapt to noisy or incomplete input data.

[0230] Preprocessing operations can be performed on the feature inputs provided to the model in box 524 to improve their utility for decision-making. These operations may include normalization to establish a consistent scale across different metrics, feature encoding to transform categorical variables into numerical representations, imputation of missing values ​​to handle incomplete telemetry, and / or dimensionality reduction to identify the main components or underlying factors capturing feature variations while reducing computational complexity. In some respects, depending on the model architecture and the nature of the relationships between different parameters, the input features may be structured into feature vectors, tensors, or graphical representations.

[0231] After processing the input features by the model in box 524, method 516 proceeds to box 526, where the model generates configuration output. Box 526 represents the generation of executable instructions that determine how to route data through the available communication interfaces of the hybrid network interface controller. The configuration output represents the optimal interface allocation determined by the model based on the provided workload and media information.

[0232] In some respects, configuration outputs can specify: interface selection instructions that determine which communication paths should be activated or deactivated (e.g., "use optical interface only", "use wireless interface only", or "concurrently enable both interfaces"); routing policies that control traffic distribution on active interfaces (e.g., application-based routing, flow-based separation, or load balancing percentage); operating parameters that tune interface behavior (e.g., transmit power level, packet size, or protocol timing values); or state transition instructions that establish conditions for dynamic reconfiguration (e.g., failover thresholds, hysteresis bands, or stability timers). Configuration outputs can also include scores or uncertainty estimates reflecting the model's confidence in its recommendations, enabling risk-aware decision-making in ambiguous scenarios.

[0233] The configuration output generated at box 526 can be represented in various formats, including but not limited to: binary flags indicating interface activation status, enumerated values ​​representing operating modes, floating-point parameters specifying resource allocation ratios, structured objects encapsulating multiple configuration dimensions, or hierarchical strategies combining high-level instructions with low-level parameter settings. The format may be determined by the model architecture, the capability requirements of downstream components, or the complexity of the configuration space being processed.

[0234] After the model generates the configuration output in box 526, method 516 proceeds to box 528, where the system provides the configuration output to the controller. Box 528 represents the transfer of the decision result from the model to the execution entity responsible for implementing the selected configuration. Depending on the system architecture, the controller mentioned in box 528 can include various components, including but not limited to: a local controller integrated within the network interface controller (e.g., refer to...). Figure 1A-2D The controller 112 described is a host system processor that manages peripheral devices or an external orchestration agent that coordinates network resources across multiple devices.

[0235] The delivery mechanism employed in box 528 can vary based on the relationship between the model and the controller. In a configuration where the model runs within the same processing domain as the controller, configuration output can be transmitted via direct function calls, shared memory structures, or internal message queues. In a distributed architecture where the model runs on a system separate from the controller, configuration output can be transmitted via network communication protocols, remote procedure calls, or management interfaces coupled to different management domains. The delivery mechanism can include reliability features, such as acknowledgment mechanisms, delivery acknowledgments, or idempotent operations, to ensure the correct receipt and application of configuration commands.

[0236] Method 516 ends in box 530, which signifies the completion of the current configuration generation cycle. In some implementations, reaching box 530 may trigger a delay interval before restarting the method in box 518, thus establishing a periodic evaluation pattern. Alternatively, the system may remain in box 530 until an event-based trigger initiates a new evaluation cycle, such as detecting changed network conditions, receiving new workload information, or an explicit reconfiguration command. The frequency of configuration updates can be tuned based on factors such as network condition volatility, workload characteristics diversity, or computational overhead associated with model evaluation.

[0237] In some embodiments, Figure 5B The model-based configuration generation process illustrated in the figure can be compared with threshold-based methods (such as references). Figure 5AThese methods operate concurrently, thereby providing complementary decision-making mechanisms for different operational scenarios or time scales. The integration of these methods can leverage techniques such as: decision fusion, where outputs from multiple decision systems are combined according to confidence weights; specialized mapping, which guides different categories of decisions to the most appropriate approach; or hierarchical organization, where one approach provides high-level instructions while another handles the detailed implementation details.

[0238] Now for reference Figure 5C The figure illustrates a flowchart of an exemplary method 532 for applying configuration output to a hybrid network interface controller according to at least one aspect of the present disclosure. Method 532 represents a process for implementing configuration instructions within a hybrid network interface device, such as the hybrid network interface controller described with respect to the previous figures. Method 532 can be implemented by a controller (such as a reference...) Figure 1A-2D The controller 112 described may execute the command, or the configuration application logic residing within the hybrid network interface controller may execute the command.

[0239] Method 532 begins in a start box 534, which represents the initialization of the configuration application process. The start box 534 can be triggered by various events, including but not limited to: receiving a new configuration instruction, detecting a configuration update notification, the expiration of a configuration refresh timer, or activating a configuration change processor in response to a system event. In some aspects, the start box 534 may include preparatory operations, such as acquiring a configuration lock to prevent concurrent modification, initializing state variables used during the configuration application, or establishing transaction boundaries for implementing configuration changes and / or rollbacks of configuration changes. The start box 534 may also include verification steps as one or more prerequisites for configuration modifications, such as verifying the operational status of the affected components, checking resource availability, or confirming authorization for the requested changes.

[0240] From the start box 534, method 532 proceeds to box 536, where the system receives configuration output. Box 536 indicates the acquisition of configuration instructions that specify the desired operating parameters, interface selection, or routing behavior of the hybrid network interface controller. The configuration output received in box 536 can originate from various sources, including but not limited to: references Figure 5B The aforementioned model-based configuration generator, external orchestration system (such as a software-defined networking (SDN) controller), network management platform, policy engine, or host system driver or management agent. Depending on the system architecture and the relationship between the configuration generator and consumers, configuration output can be delivered through various mechanisms, including control plane messages, management interface commands, memory-mapped registers, or inter-process communication channels.

[0241] In some aspects, the configuration output may include structured data specifying operational instructions for one or more components of the hybrid network interface controller. This structured data may include parameters such as interface activation indicators specifying which communication interfaces should be enabled or disabled (e.g., optical active, wireless inactive), traffic routing rules determining how data streams should be distributed across available interfaces (e.g., high-priority traffic via optical, background transmission via wireless), performance parameters controlling interface behavior (e.g., transmission power levels, protocol timing values, buffer allocation), and / or state transition instructions for dynamic reconfiguration setup conditions (e.g., failover thresholds, stability timers). The configuration output may be encoded in various formats, including binary control structures, JSON or XML documents, key-value pairs, protocol buffers, or proprietary representation formats optimized for efficient processing or minimized storage requirements.

[0242] The receive operation in box 536 may include preprocessing steps such as decoding the serialized configuration representation, validating the configuration syntax and semantics, resolving dependencies between configuration elements, or converting high-level instructions into device-specific settings. In some implementations, the receive operation may incorporate filtering or transformation logic that adapts generic configuration instructions to the specific capabilities and constraints of the current hardware platform, thereby ensuring portability across different device generations or variants.

[0243] After receiving the configuration output in box 536, method 532 proceeds to box 538, where the system applies the configuration. Box 538 represents the execution phase, during which the configuration instructions are translated into operational state changes within the hybrid network interface controller. Depending on the nature and scope of the configuration instructions, application operations can include a variety of actions, including but not limited to: modifying register values ​​that control the behavior of hardware components, updating routing tables or flow classifiers that determine data path selection, adjusting buffer allocations or quality of service parameters that affect performance characteristics, or reconfiguring the protocol engine that implements the communication standard.

[0244] In some respects, the application operations in Box 538 can be organized as a multi-phase process that ensures consistency and minimizes service disruptions during configuration changes. This process may include a preparation phase that allocates and verifies resources before modification, an execution phase that performs the actual state changes, and a verification phase that tests the correctness and stability of the applied configuration. Application operations may incorporate transactional semantics, which implement atomic applications of the associated configuration changes, thereby ensuring that the system fully applies a consistent set of configurations or retains its previous state even if errors occur during application.

[0245] For interface-specific configurations, application operations can include hardware-level actions such as enabling or disabling power to transmit or receive circuits, configuring physical layer parameters such as modulation schemes or signal levels, or updating media access control settings that control protocol behavior. For routing configurations, application operations can modify packet bootstrapping rules, update flow tables, or reconfigure switching elements to bootstrap traffic based on specified policies or workload characteristics. For adaptive configurations, application operations can adjust thresholds based on operating conditions, monitoring parameters, or decision criteria used by control algorithms that dynamically optimize interface selection.

[0246] Method 532 concludes in box 540, which signifies the completion of the application configuration process. In some implementations, reaching box 540 may trigger a notification mechanism that informs interested components or systems of the configuration changes, allowing them to adjust their behavior accordingly. Such notifications may be delivered to host systems to update their understanding of available communication paths, to performance monitoring systems to establish appropriate baseline measurements of the new configuration, or to higher-level orchestration systems to confirm the successful application of the requested settings. The completion phase may also include cleanup operations such as releasing configuration locks, finalizing transaction logs, or updating configuration version indicators to track applied changes for auditing or rollback purposes.

[0247] Now for reference Figure 6 The present disclosure describes a flowchart of a method 600 for dynamically selecting between an optical communication interface and a wireless communication interface using AI-based decision logic, wherein at least some concepts have been discussed with reference to Figures 1-5C. Method 600 can be derived from... Figure 1A and Figure 1B The hybrid network interface controller 100A / 100B described herein is executed by one or more control elements (such as controller 112, communication medium selector 202A / 202B, and model 204). In one aspect, method 600 can be implemented in one or more circuits. As will be understood, one or more circuits can be included Figure 2A and Figure 2B In the controller 112, the communication medium selector 202A / 202B, and / or corresponding to or involved in Figure 4D The elements described in the hybrid optical transmission system 490.

[0248] Method 600 begins in block 602, which represents the start condition for the data routing process. The method proceeds to block 604, in which the hybrid network interface device, in response to a threshold condition, selectively routes data from the host device to at least one of an optical interface or a wireless interface. That is, the hybrid network interface device can dynamically determine which communication medium to use based on various operational parameters that can trigger the threshold condition.

[0249] The threshold conditions mentioned in box 604 can be based on a variety of factors, including but not limited to the measured communication link-quality metrics, bandwidth utilization, detected bit error rate, or signal-to-noise ratio associated with optical and wireless communication media. In some respects, threshold conditions can include workload requirements, traffic priority levels, or can be determined by analyzing usage metrics collected from both interfaces using AI-based decision logic.

[0250] After completing the selective routing operation, method 600 proceeds to box 606, which indicates the end of the current routing decision cycle.

[0251] In some aspects of method 600, the one or more circuits are also used to: switch data transmission from a first communication medium to a second communication medium; or concurrently enable both communication media.

[0252] In some aspects of method 600, the one or more circuits are also used to route data via optical and wireless interfaces in response to threshold conditions.

[0253] In some aspects of method 600, the threshold condition is based on at least one of the following: a measured communication link quality metric associated with at least one of the optical communication medium or the wireless communication medium; a measurement of bandwidth utilization associated with at least one of the optical communication medium or the wireless communication medium; a detected bit error rate associated with at least one of the optical communication medium or the wireless communication medium; or a signal-to-noise ratio associated with at least one of the optical communication medium or the wireless communication medium.

[0254] In some aspects of method 600, the one or more circuits are further configured to: receive an indication of workload demand; and compare the indication of workload demand with the threshold condition.

[0255] In some aspects of method 600, the one or more circuits are further configured to: route high-priority workload traffic via the optical interface; and route lower-priority workload traffic via the wireless interface.

[0256] In some aspects of method 600, the one or more circuits include a local memory storing artificial intelligence (AI)-based decision logic, and the threshold condition is determined by the AI-based decision logic in response to usage metrics collected from the optical interface and the wireless interface.

[0257] In some aspects of method 600, the threshold condition is triggered in response to receiving a switching command from an external orchestration controller in the data center network.

[0258] In some aspects of method 600, the network interface device may include: a coupling interface for coupling the network interface device to the host device, wherein the coupling interface is further configured to communicate with the host device using at least one of an optical path or an electrical path.

[0259] In some aspects of method 600, the wireless interface includes at least one of the following: a photonic mixer stage for generating optical signals; or a radio frequency mixer stage configured to generate or receive millimeter-wave or terahertz signals.

[0260] In some aspects of method 600, the network interface device may include: a photonic signal generator for generating a modulated optical signal; and a branching module for directing the modulated optical signal to at least one of the optical interface or the wireless interface.

[0261] In some aspects of method 600, the network interface device may include an optical output that couples the modulated optical signal to an optical fiber.

[0262] In some aspects of method 600, the wireless interface includes: a photonic mixer stage for generating millimeter-wave or terahertz signals; and an antenna coupled to the photonic mixer stage.

[0263] In view of the foregoing, it should be understood that one aspect of this disclosure provides a network interface device comprising: an optical interface for receiving a modulated optical signal; an optical processing circuit for demodulating the modulated optical signal; a wireless interface for receiving a wireless signal; a wireless processing circuit for decoding the wireless signal; and one or more output circuits coupled to the optical processing circuit and the wireless processing circuit, the one or more output circuits being used to provide demodulated data to a host system. In some aspects, the network interface device includes a controller communicatively coupled to the optical processing circuit and the wireless processing circuit, the controller being configured to: monitor at least one link quality metric associated with the optical interface and the wireless interface; and selectively enable or disable reception through at least one of the optical interface or the wireless interface based on a threshold condition. In some aspects, the controller is further configured to simultaneously receive data through the optical interface and the wireless interface in response to detecting that at least one of total bandwidth requirement or link utilization meets or exceeds the threshold condition. In some aspects, the controller includes artificial intelligence (AI) based decision logic for modifying the threshold condition based on the at least one link-quality metric.

[0264] In view of the foregoing, it should be understood that aspects of this disclosure provide a system comprising: a host device for generating outgoing data traffic; a network interface device communicatively coupled to the host device, the network interface device comprising: an optical interface for transmitting data via an optical communication medium, a wireless interface for transmitting data via a wireless communication medium, and a controller for selectively routing the outgoing data traffic between the optical interface and the wireless interface based on threshold conditions. In some aspects, the network interface device further comprises: optical processing circuitry for demodulating incoming optical signals from the optical communication medium; and wireless processing circuitry for decoding incoming wireless signals from the wireless communication medium, wherein the controller is further configured to route incoming data to the host device. In some aspects, the controller includes an artificial intelligence module for: monitoring link performance metrics from the optical interface and the wireless interface; and dynamically adjusting the threshold conditions in response to changes in the link performance metrics. In some aspects, the host device includes a workload manager that assigns priority indicators to outgoing data, and the controller is configured to route higher-priority traffic via the optical interface and lower-priority traffic via the wireless interface.

[0265] Figure 7A The illustration shows a block diagram of a system 700 according to at least one embodiment, including a switching circuit 710 (sometimes referred to as a "structure") located between a first group 712 devices and a second group 714 devices. The switching circuit 710 may be a hybrid switch supporting wired and wireless communication (e.g., Figure 7B The system 700 can be implemented using a switch 756. As a non-limiting example, the system 700 can be implemented as a GPU-to-GPU link system (e.g., an NVLINK® GPU-to-GPU interconnect system), where the switching circuitry 710 may include one or more GPU-to-GPU switches (e.g., one or more NVSWITCH™ switches). The system 700 can also be implemented as other systems (e.g., as a component of such systems), such as data centers, cloud computing systems, machine learning systems (e.g., utilizing one or more neural networks), autonomous vehicles, medical imaging equipment, etc.

[0266] Each of the first group 712 and the second group 714 can be implemented as a GPU, a parallel processing unit, a central processing unit (“CPU”), a data processing unit (“DPU”), a controller, a switch, a switching circuit, a storage device, etc. As a non-limiting example, the first group 712 has been illustrated as including several (“M”) devices illustrated as SD1-1 to SDM-X. X” devices. However, group 712 may include any number of devices, including a single device. For ease of illustration, in Figure 7AIn the diagram, devices SD1-1 to SDM-X are illustrated as a two-dimensional array comprising "M" rows and "X" columns. The second group 714 is illustrated as comprising "Y" devices, illustrated as devices T1 to TY, but the second group 714 may include any number of devices, including a single device. Furthermore, one or more devices may be members of both groups 712 and 714. Therefore, groups 712 and 714 are not mutually exclusive and may include one or more of the same devices. Thus, in at least one embodiment, the switching circuit 710 may allow any device in the first group 712 (SD1-1 to SDM-X) to communicate with any device in the first group 712 (SD1-1 to SDM-X) and / or any device in the second group 714 (T1 to TY). Therefore, in at least one embodiment, the switching circuit 710 may allow any of the devices T1 to TY in the second group 714 to communicate with any of the devices T1 to TY in the second group 714 and / or any of the devices SD1-1 to SDM-X in the first group 712.

[0267] Switching circuit 710 may include an N number of outbound switches OSW1 to OSWN, and a predetermined number of M switches. The internal switches IS1C1 to ISMCN of “N”, signal conductor 716, and routing block 718 controlling the operation of internal switches IS1C1 to ISMCN. Routing block 718 has been illustrated to include hardware 720, which may include at least one processor 722, memory 724, and one or more counters 726. In at least one embodiment, each of the one or more processors 722 may be implemented as one or more hardware state machines, one or more microprocessors, one or more microcontrollers, one or more controllers, etc. Memory 724 (e.g., one or more non-transitory processor-readable media) may be implemented, for example, using volatile memory (e.g., dynamic random access memory (“DRAM”) and / or non-volatile memory (e.g., hard disk drives, solid-state devices (“SSDs”), etc.).

[0268] Signal conductors 716 interconnect the first group 712, the second group 714, the internal switches IS1C1 to ISMCN, and the outbound switches OSW1 to OSWN. In the illustrated example, signal conductors 716 may include any number of signal conductors. Each of the signal conductors 716 may be implemented as a wire, a trace, and / or any other type of electrical conductor. As a non-limiting example, one or more of the signal conductors 716 may each be implemented as a wire-based serial short-range communication link. As another non-limiting example, one or more of the signal conductors 716 may each be implemented as a Scalable Link Interface (“SLI”).

[0269] In the illustrated embodiment, signal conductor 716 includes signal conductors 734, 740, and 742. Signal conductor 734 connects devices SD1-1 to SDM-X to internal switches IS1C1 to ISMCN. For ease of illustration, Figure 7A Only the signal conductors 734 connecting each of devices SD1-1 to SD1-X to internal switches IS1C1 to IS1CN are illustrated. In at least one embodiment, internal switches IS1C1 to IS1CN may each include a different port connected via signal conductors 734 to at least a portion of devices SD1-1 to SDM-X (e.g., connected to devices SD1-1 to SD1-X respectively). Similarly, internal switches IS2C1 to IS2CN may each include a different internal port connected via signal conductors 734 to at least a portion of devices SD1-1 to SDM-X (e.g., connected to devices SD2-1 to SD2-X respectively). Further, internal switches ISMC1 to ISMCN may each include a different internal port connected via signal conductors 734 to at least a portion of devices SD1-1 to SDM-X (e.g., connected to devices SDM-1 to SDM-X respectively).

[0270] Signal conductors 742 connect internal switches IS1C1 to ISMCN to outbound switches OSW1 to OSWN. For example, at least one signal conductor in signal conductors 742 connects each internal switch from IS1C1 to ISMC1 to outbound switch OSW1, at least one signal conductor in signal conductors 742 connects each internal switch from IS1C2 to ISMC2 to outbound switch OSW2, and at least one signal conductor in signal conductors 742 connects each internal switch from IS1CN to ISMCN to outbound switch OSWN.

[0271] Signal conductor 740 connects outbound switches OSW1 to OSWN to devices T1 to TY. Outbound switches OSW1 to OSWN may each include different ports, which are connected to at least one device among devices T1 to TY via at least one signal conductor in signal conductor 740.

[0272] Although switching circuit 710 has been illustrated as including internal switches IS1C1 to ISMCN, switching circuit 710 may include any number of internal switches, including a single internal switch. Further, internal switches IS1C1 to ISMCN have been illustrated as a two-dimensional array comprising "M" rows and "N" columns. However, this is not required, and internal switches IS1C1 to ISMCN can be positioned in alternative arrangements to achieve any network topology, including a three-dimensional array. One or more internal switches IS1C1 to ISMCN may be coupled to each other. For example, from the perspective of one internal switch IS1C1 to ISMCN, one or more of the other internal switches IS1C1 to ISMCN may be one of devices T1 to TY. Further, switching circuit 710 may be coupled to one or more switches external to switching circuit 710 (e.g., a switching circuit system such as switching circuit 710). For example, one or more devices SD1-1 to SDM-X and / or devices T1 to TY may each be implemented as an external switch.

[0273] The routing block 718 can be connected via one or more buses to each of the internal switches IS1C1 to ISMCN and / or at least a portion of each of the signal conductors 716. For example, one or more processors 722 and memory 724 can be connected to each other via one or more buses 728. One or more buses 728 can also connect one or more processors 722 and memory 724 to each of the internal switches IS1C1 to ISMCN and / or at least a portion of each of the signal conductors 716. Memory 724 (e.g., a non-transitory processor-readable medium) stores instructions 730, data 732, and firmware 752 executable by one or more processors 722, which can perform monitoring and data manipulation functions. For example, firmware 752 can instruct hardware 720 to transition one or more of the signal conductors 716 from an active state to an inactive state, and vice versa.

[0274] Figure 7BA hybrid switch 756, as described in more detail herein, is depicted that enables communication between processing devices and / or client devices. Switch 756 is capable of receiving and transmitting data and can act as a central node in a network. Switch 756 can be cabled in a topology including spine switches and top-of-rack (TOR) switches. Switch 756 is capable of receiving, processing data (e.g., packets), and forwarding them to appropriate destinations within the network. In some aspects, such destinations may include, but are not limited to, processing devices and / or client devices. In some implementations, as described herein, switch 756 may be included in a switch box, platform, or enclosure that may include one or more switches 756 and one or more power supply switch boxes, platforms, or enclosures.

[0275] In some implementations, one or more processing devices may be connected to one or more ports of one or more switches 756 via network cables. In some aspects, one or more processing devices may be wirelessly connected to one or more ports of one or more switches 756. Processes performed by the processing devices, such as applications, may involve transmitting data to nodes in the network, such as to other processing devices and / or client devices. Data may flow through the network of processing devices and switches 756 using one or more protocols, such as Transmission Control Protocol (TCP), User Datagram Protocol (UDP), or Internet Protocol (IP). After receiving data from a processing device or another switch 756, one or more switches 756 may examine the data to identify its destination and route the data across the network. In some aspects, switches 756 may integrate features such as… Figure 1A-1B The one or more network interface controllers described.

[0276] In some implementations, the architecture may include a wireless communication device that establishes a wireless link between the processing device and / or the switch 756. The wireless communication device may utilize millimeter-wave or terahertz frequencies to provide high-bandwidth, low-latency wireless connectivity suitable for data center applications. One or more switches 756 may combine hybrid connectivity capabilities to support both wired and wireless communication interfaces. Such hybrid switches 756 may dynamically allocate communication resources based on workload priorities, bandwidth requirements, and current network conditions.

[0277] In some aspects, switch 756 may include multiple ports 758a-758n / 760a-760n, switching hardware 766, processing circuitry 770, and memory 772. Switch 756 may include one or more ports 758a-758n / 760a-760n configured to send and receive data in the form of packets, frames, and / or other information units. Ports 758a-758n / 760a-760n may be referred to as hardware ports, indicating a physical interface capable of handling communication over copper, fiber optic, or radio frequency links. Ports 758a-758n (wired) support transmission via physical cable routing, while ports 760a-760n (wireless) support wireless communication protocols suitable for high-throughput data center environments. A "hardware port" can refer to a physical interface on a network switching device (also referred to herein as a switch and network switch), such as an Ethernet or fiber optic port, through which network packets are sent and received via a physical medium. A hardware port may differ from a "logical port," which can refer to a software-defined endpoint identified by a numerical designation used to facilitate protocol-specific communication between network interconnection devices in the transport or application layer. While embodiments primarily relating to hardware ports have been discussed, embodiments of this disclosure can also be applied to logical ports.

[0278] In some respects, each 758a-758n port can support high-speed data transmission over physical cabling (such as copper or fiber optic cables). In some respects, one or more 760a-760n ports can be compatible with one or more wireless transceivers. In some respects, a 760a-760n port can refer to a physical radio interface, logical port, or SSID bound to a virtual access point, and / or a beamforming directional link to a specific destination node. Ports 758a-758n / 760a-760n can be configured to accommodate one or more communication standards and protocols, including Ethernet, InfiniBand, and / or proprietary link layer protocols, enabling flexible deployment in a variety of network interconnection environments. Depending on the overall configuration and management strategy, configured via, for example, software or firmware, each 758a-758n / 760a-760n port can dynamically switch between ingress and egress functions, or operate in a dedicated ingress or egress mode.

[0279] In some aspects, each port 758a-758n / 760a-760n may incorporate physical layer devices (PHYs) and / or logic that perform signal encoding / decoding, clocking and data recovery, and other low-level operations. These components may be implemented in an ASIC, FPGA, or discrete module. Ports may also include local buffers or queues, queues 762a-762n for wired ports and queues 764a-764n for wireless ports, to temporarily store data before switching. In some implementations, port-specific buffers or queues 762a-762n / 764a-764n may be referred to as on-chip or external memory elements. On-chip or external memory elements may work in conjunction with larger shared resources, such as buffer 768, to provide efficient queuing and flow management. The configuration and operational status of each port 758a-758n / 760a-760n, such as speed, duplex setting, or active / inactive state, may be monitored by processing circuitry 770. Furthermore, ports 758a-758n / 760a-760n can provide status indicators or counters (e.g., for transmitted or dropped packets), which can be accessed by higher-level software for analysis or troubleshooting. Ports 758a-758n / 760a-760n can provide one or more interface points for connecting switch 756 to external devices, thereby enabling one or more traffic management, congestion management, and / or load balancing techniques described herein. In some aspects, one or more queues 762a-762n / 764a-764n can be used for per-client buffering, traffic prioritization, MAC layer retransmission, and / or QoS and airtime fairness implementation.

[0280] In some aspects, one or more queues 762a-762n / 764a-764n can be configured to temporarily store data units (e.g., packets) awaiting transmission or processing. In some implementations, one or more queues 762a-762n / 764a-764n can be implemented in hardware as dedicated first-in-first-out (FIFO) buffers, circular buffers, or other structures capable of dynamic partitioning. For example, in an implementation of switch 756 that supports multiple "priority levels," high-priority traffic (such as real-time video or voice data) can be assigned to a dedicated queue with accelerated processing, while low-priority data (such as background file transfers) can be placed in another queue with a best-effort service policy.

[0281] The number of queues 762a-762n / 764a-764n can vary depending on one or more configurations. For example, one implementation may assign a separate queue to each port 758a-758n / 760a-760n. In other implementations, one or more queues 762a-762n / 764a-764n may be allocated based on traffic category, "priority level," or virtual channel. In some respects, one or more queues 762a-762n / 764a-764n collectively perform flow control tasks by absorbing surges in incoming data and releasing them at a rate consistent with downstream capacity. For example, in a data center environment running HPC applications, each queue 762a-762n / 764a-764n may correspond to a specific "traffic type," such as compute, storage, or management, and selectively throttle certain types of traffic whenever there is a rapid increase in traffic across multiple ports simultaneously. By grouping traffic in this way, the system can handle surges in new data deliveries without saturating downstream resources.

[0282] In some aspects, the switching hardware 766 may include one or more integrated circuits, multiplexers, cross-connects, or other data forwarding components arranged to route incoming data from ports (e.g., one or more ports 758a-758n / 760a-760n) to one or more destination ports based on switching logic for Layer 2 forwarding and / or routing logic for Layer 3 processing. In some implementations, the switching hardware 766 may operate at line rate (i.e., process packets at the maximum theoretical speed of the network interface) to ensure minimal latency while maintaining reliable throughput. The switching hardware 766 may incorporate dedicated forwarding tables or pipelines that analyze header information, such as Media Access Control (MAC) addresses for Layer 2 switching, Virtual LAN (VLAN) tags, or network layer identifiers (such as Internet Protocol (IP)) for Layer 3 routing, to determine how best to forward each data unit within the switch 756. Furthermore, the switching hardware 766 can utilize programmable matching action levels (i.e., a processing pipeline level that compares incoming packet fields with predefined patterns and performs corresponding actions), enabling it to adapt to evolving network requirements through software-defined updates. For example, reconfiguring the switching hardware 766 can implement different flow control policies, advanced monitoring capabilities, or security filters without requiring a complete redesign of the switching hardware 766.

[0283] Processing circuitry 770 may include one or more processors, microcontrollers, digital signal processors (DSPs), or dedicated ASIC blocks configured to manage system-level operations of switch 756. In some implementations, processing circuitry 770 may execute firmware stored in memory 772 to maintain routing tables, management port configuration parameters (e.g., enabling or disabling specific ports 758a-758n / 760a-760n), and apply quality of service (QoS) policies. Processing circuitry 770 may interface with higher-level management tools, such as network orchestration platforms or cloud-based controllers, to transmit telemetry data in real time or receive updated resource management circuitry / logic. By offloading some decisions to the switching hardware 766, processing circuitry 770 remains available for more complex tasks, including congestion analysis, device configuration, and other diagnostic tasks (e.g., performing flow control counter debugging or verifying QoS compliance). Therefore, switch 756 can maintain high levels of throughput while still providing congestion management and adaptive routing capabilities.

[0284] Switch 756 may include memory 772, which may utilize one or more storage technologies, such as DRAM, SRAM, TCAM, or flash memory. In some aspects, DRAM may store large routing tables or buffered data, SRAM may serve as a high-speed temporary register for real-time operation, and TCAM may facilitate fast lookups for routing or security rules. Regardless of the specific technology, memory 772 may be used to make processing circuitry 770 easily accessible to configuration parameters, firmware images, and runtime data structures. In some aspects, memory 772 may house a lightweight operating system or embedded software stack that manages hardware resources and provides an interface for remote configuration and updates. Furthermore, memory 772 may implement virtualization capabilities, allowing multiple logical partitions (or “slices”) of the device to operate independently, which can be useful in large-scale data center or cloud environments where different tenants or workloads may require isolated resources and custom flow control settings.

[0285] Figure 8 A block diagram of an example system 800 according to at least one embodiment is illustrated. In at least one embodiment, system 800 may include one or more CPUs 810 connected to one or more multi-GPU communication systems (e.g., multi-GPU communication system 802) via one or more signal conductors 812. Figure 8In the embodiment illustrated, one or more CPUs 810 include CPUs 810A, 810B, 810C, and 810D respectively connected to the multi-GPU communication system 802 via signal conductors 812A, 812B, 812C, and 812D. In some aspects, zero, one, or more CPUs (e.g., 810N) may communicate with a switch (e.g., 806D) via one or more signal conductors (e.g., 812N). Figure 8 In the multi-GPU communication system 802, GPUs 808 are interconnected via at least one switch 806 and at least one signal conductor 804. In the illustrated embodiment, one or more switches 806 include switches 806A-806D, GPUs 808 include GPUs 808A-808P, and one or more signal conductors 804 include signal conductors 804A-804E.

[0286] System 800 can be characterized as an example implementation of System 700 (see [link]). Figure 7A For example, one or more CPUs 810 and / or GPUs 808 may be implemented as groups 712 and / or 714. In at least one embodiment, one or more CPUs 810 may be substantially the same as devices SD1-1 to SDM-X and / or devices T1 to TY. In at least one embodiment, GPUs 808 may be substantially the same as devices SD1-1 to SDM-X and / or devices T1 to TY. As another non-limiting example, one or more signal conductors 804 and / or one or more signal conductors 812 may be implemented as signal conductors 716. In at least one embodiment, one or more signal conductors 804 and / or one or more signal conductors 812 may be substantially the same as signal conductors 716. As yet another non-limiting example, one or more switches 806 may be implemented as internal switches IS1C1 to ISMCN and / or outbound switches OSW1 to OSWN. In at least one embodiment, one or more switches 806 may be implemented as switches IS1C1 to ISMCN (see... Figure 7A The outbound switches OSW1 to OSWN are basically the same.

[0287] As a non-limiting example, signal conductor 804 can link GPU 808 to one or more switches 806, and signal conductors 812A and 812B can link at least one of the switches 806 to one or more CPUs 810. Figure 8As illustrated, CPUs 810A and 810B can be connected to switches 806A and 806C respectively via signal conductors 812A and 812B. In at least one embodiment, one or more of the CPUs 810 can be implemented as follows: Figure 8 The CPU and GPU 808 illustrated can be implemented as one or more GPUs. Figure 14 The illustration shows one or more parallel processors 1412.

[0288] The multi-GPU communication system 802 can be implemented as a mesh network interconnect, which can have a local area network topology in which infrastructure nodes or devices can be directly, dynamically, and non-hierarchically connected to other nodes and can cooperate with each other to efficiently route data. Examples of infrastructure devices may include bridges, one or more switches 806, and / or other types of infrastructure devices. The multi-GPU communication system 802 may be referred to as a GPU-to-GPU system. By way of non-limiting example, the multi-GPU communication system 802 can be implemented as a GPU-to-GPU link system (e.g., an NVLINK® GPU-to-GPU interconnect system), in which one or more switches 806 may include one or more GPU-to-GPU switches (e.g., one or more NVSWITCH™ switches).

[0289] Switch 806A is connected to GPUs 808A-808D via signal conductor 804A, switch 806B is connected to GPUs 808E-808H via signal conductor 804B, switch 806C is connected to GPUs 808I-808L via signal conductor 804C, and switch 806D is connected to GPUs 808M-808P via signal conductor 804D. Switches 806A-806D are interconnected via signal conductor 804E. During the duration of the dedicated connection, each of the one or more switches 806 can act as a connection between two endpoints within a router and / or system 800. In at least one embodiment, one or more switches of the one or more switches 806 can be implemented using a node-on-node switch architecture (e.g., one or more NVSWITCH™ switches) that connects processing units (e.g., one or more GPUs, one or more CPUs, and / or one or more other types of processors). System 800 can be used as part of another system (such as a data center) that includes hardware, network, software and / or libraries, and executes applications that perform inference, for example, by executing artificial intelligence (“AI”) or machine learning models.

[0290] The systems and methods described herein can be used, but are not limited to, non-autonomous vehicles, semi-autonomous vehicles (e.g., in one or more adaptive driver assistance systems (ADAS)), manned and unmanned robots or robotic platforms, warehouse vehicles, off-road vehicles, vehicles coupled to one or more trailers, aircraft, boats, shuttles, emergency response vehicles, motorcycles, electric or motorized bicycles, aircraft, engineering vehicles, underwater vehicles, drones, and / or other types of vehicles. The systems and methods described herein can be used in augmented reality, virtual reality, mixed reality, robotics, security and supervision, autonomous or semi-autonomous machine applications, and / or any other technological space where one or more signal conductors may have at least two distinct states consuming different amounts of power.

[0291] The systems and methods described herein can be used, but are not limited to, non-autonomous vehicles, semi-autonomous vehicles (e.g., in one or more adaptive driver assistance systems (ADAS)), manned and unmanned robots or robotic platforms, warehouse vehicles, off-road vehicles, vehicles coupled to one or more trailers, aircraft, boats, shuttles, emergency response vehicles, motorcycles, electric or motorized bicycles, aircraft, engineering vehicles, underwater vehicles, drones, and / or other types of vehicles. Furthermore, the systems and methods described herein can be used for a variety of purposes, by way of example but not limited to, machine control, machine motion, machine driving, synthetic data generation, model training, perception, augmented reality, virtual reality, mixed reality, robotics, safety and supervision, simulation and digital twins, autonomous or semi-autonomous machine applications, deep learning, environmental simulation, object or actor simulation and / or digital twins, data center processing, conversational AI, optical transport simulation (e.g., ray tracing, path tracing, etc.), collaborative content creation for 3D assets, cloud computing, web hosting services or web hosting platforms, and / or any other suitable applications.

[0292] The disclosed embodiments can comprise a variety of different systems, such as automotive systems (e.g., control systems for autonomous or semi-autonomous machines, perception systems for autonomous or semi-autonomous machines), systems implemented using robots, aviation systems, medical systems, boating systems, smart area monitoring systems, systems for performing deep learning operations, systems for performing simulation operations, systems for performing digital twin operations, systems implemented using edge devices, systems containing one or more virtual machines (VMs), systems for performing synthetic data generation operations, systems implemented at least partially in a data center, systems for performing conversational AI operations, systems for performing optical transmission simulations, systems for performing collaborative content creation for 3D assets, systems implemented at least partially using cloud computing resources, systems for implementing web hosting services (e.g., runtime for program optimization) or web hosting platforms (e.g., integrated development environments including program optimization as a service) (as an application programming interface (“API”) between two or more separate applications or systems), and other types of systems.

[0293] Data Center Figure 9 An example data center 900 according to at least one embodiment is shown. In at least one embodiment, the data center 900 includes, but is not limited to, a data center infrastructure layer 910, a framework layer 920, a software layer 930, and an application layer 940.

[0294] In at least one embodiment, such as Figure 9 As shown, the data center infrastructure layer 910 may include a resource coordinator 912, grouped computing resources 914, and node computing resources (“nodes CR”) 916(1)-916(N), where “N” represents any complete positive integer. In at least one embodiment, nodes CR 916(1)-916(N) may include, but are not limited to, any number of central processing units (“CPUs”) or other processors (including accelerators, field-programmable gate arrays (“FPGAs”), data processing units (“DPUs”) in network devices, graphics processors, etc.), memory devices (e.g., dynamic read-only memory), storage devices (e.g., solid-state drives or disk drives), network input / output (“NW I / O”) devices, network switches, virtual machines (“VMs”), power modules, and cooling modules, etc. In at least one embodiment, one or more of nodes CR 916(1)-916(N) may be servers having one or more of the aforementioned computing resources.

[0295] In at least one embodiment, the grouped computing resources 914 may include individual groups (not shown) of node CRs housed in one or more racks, or a plurality of racks (also not shown) housed in data centers in various geographic locations. The individual groups of node CRs within the grouped computing resources 914 may include computing, networking, memory, or storage resources that can be configured or allocated to support groups of one or more workloads. In at least one embodiment, several node CRs, including CPUs or processors, may be grouped within one or more racks to provide computing resources to support one or more workloads. In at least one embodiment, one or more racks may also include any number of power modules, cooling modules, and network switches, in any combination.

[0296] In at least one embodiment, resource coordinator 912 may be configured or otherwise control one or more nodes CR 916(1)-916(N) and / or grouped computing resources 914. In at least one embodiment, resource coordinator 912 may include a Software Design Infrastructure (“SDI”) management entity for data center 900. In at least one embodiment, resource coordinator 912 may include hardware, software, or some combination thereof.

[0297] In at least one embodiment, such as Figure 9As shown, the framework layer 920 includes, but is not limited to, a job scheduler 932, a configuration manager 934, a resource manager 936, and a distributed file system 938. In at least one embodiment, the framework layer 920 may include a framework of software 952 supporting the software layer 930 and / or one or more applications 942 of the application layer 940. In at least one embodiment, the software 952 or application 942 may respectively include web-based service software or applications, such as services or applications provided by Amazon Web Services, Google Cloud, and Microsoft Azure. In at least one embodiment, the framework layer 920 may be, but is not limited to, a free and open-source software web application framework, such as Apache Spark™ (hereinafter referred to as "Spark") which can leverage the distributed file system 938 for large-scale data processing (e.g., "big data"). In at least one embodiment, the job scheduler 932 may include a Spark driver to facilitate the scheduling of workloads supported by the various layers of the data center 900. In at least one embodiment, the configuration manager 934 may be able to configure different layers, such as the software layer 930 and the framework layer 920 including Spark and the distributed file system 938 for supporting large-scale data processing. In at least one embodiment, resource manager 936 is capable of managing cluster or group computing resources mapped to or allocated to support distributed file system 938 and job scheduler 932. In at least one embodiment, cluster or group computing resources may include group computing resources 914 on data center infrastructure layer 910. In at least one embodiment, resource manager 936 may coordinate with resource coordinator 912 to manage these mapped or allocated computing resources.

[0298] In at least one embodiment, the software 952 included in the software layer 930 may include software used by at least a portion of the nodes CR 916(1)-916(N), the grouped computing resources 914, and / or the distributed file system 938 of the framework layer 920. One or more types of software may include, but are not limited to, Internet web page search software, email virus scanning software, database software, and streaming video content software.

[0299] In at least one embodiment, one or more applications 942 included in the application layer 940 may include one or more types of applications used by at least a portion of nodes CR 916(1)-916(N), grouped computing resources 914, and / or the distributed file system 938 of the framework layer 920. One or more types of applications may include, but are not limited to, CUDA applications.

[0300] In at least one embodiment, any of the configuration manager 934, resource manager 936, and resource coordinator 912 can implement any number and type of self-modification actions based on any amount and type of data acquired in any technically feasible manner. In at least one embodiment, self-modification actions can mitigate potentially poor configuration decisions by data center operators of data center 900 and can prevent underutilization and / or poor performance of the data center.

[0301] In at least one embodiment, data center 900 can be used to implement system 700 (see...). Figure 7A ) and / or System 800 (see Figure 8 For example, the first group 712 and / or the second group 714 may include one or more groups of computing resources 914 and / or one or more CRs 916(1)-916(N). In at least one embodiment, Figure 9 One or more systems depicted in Figure 1 are used to achieve, for example, in combination with Figure 1- Figure 8 The description refers to one or more systems and / or processes of the like.

[0302] Computer-based systems The following figures illustrate exemplary computer-based systems that can be used to implement at least one embodiment, but are not limited thereto.

[0303] Figure 10 The figure illustrates a processing system 1000 according to at least one embodiment. In at least one embodiment, the processing system 1000 includes one or more processors 1002 and one or more graphics processors 1008, and may be a single-processor desktop system, a multi-processor workstation system, or a server system having a large number of processors 1002 or processor cores 1007. In at least one embodiment, the processing system 1000 is a processing platform integrated in a system-on-a-chip (“Sort”) integrated circuit for use in mobile, handheld, or embedded devices.

[0304] In at least one embodiment, the processing system 1000 may include, or be integrated with, a server-based game platform, game console, media console, mobile game console, handheld game console, or online game console. In at least one embodiment, the processing system 1000 is a mobile phone, smartphone, tablet computing device, or mobile internet device. In at least one embodiment, the processing system 1000 may also include, be coupled to, or be integrated into a wearable device, such as a smartwatch, smart glasses, augmented reality, or virtual reality device. In at least one embodiment, the processing system 1000 is a television or set-top box device having one or more processors 1002 and a graphical interface generated by one or more graphics processors 1008.

[0305] In at least one embodiment, one or more processors 1002 each include one or more processor cores 1007 for processing instructions that, when executed, perform operations for system and user software. In at least one embodiment, each of the one or more processor cores 1007 is configured to process a particular instruction set 1009. In at least one embodiment, the instruction set 1009 may facilitate Complex Instruction Set Computing (“CISC”), Reduced Instruction Set Computing (“RISC”), or computation via Very Long Instruction Word (“VLIW”). In at least one embodiment, the processor cores 1007 may each process a different instruction set 1009, which may include instructions that facilitate the emulation of other instruction sets. In at least one embodiment, the processor cores 1007 may also include other processing devices, such as digital signal processors (“DSPs”).

[0306] In at least one embodiment, processor 1002 includes cache memory ('cache') 1004. In at least one embodiment, processor 1002 may have a single internal cache or multiple levels of internal cache. In at least one embodiment, the cache memory is shared among various components of processor 1002. In at least one embodiment, processor 1002 also uses external caches (e.g., level 3 ("L3") caches or last-level caches ("LLC")) (not shown), which can be shared among processor cores 1007 using known cache coherence techniques. In at least one embodiment, processor 1002 also includes a register file 1006, which may include different types of registers (e.g., integer registers, floating-point registers, status registers, and instruction pointer registers) for storing different types of data. In at least one embodiment, register file 1006 may include general-purpose registers or other registers.

[0307] In at least one embodiment, one or more processors 1002 are coupled to one or more interface buses 1010 to transmit communication signals, such as address, data, or control signals, between the processors 1002 and other components in the processing system 1000. In at least one embodiment, the interface bus 1010 may be a processor bus, such as a version of a Direct Media Interface (“DMI”) bus. In at least one embodiment, the interface bus 1010 is not limited to a DMI bus and may include one or more peripheral component interconnect buses (e.g., “PCI”, “Fast PCI (“PCIe”)”), memory buses, or other types of interface buses. In at least one embodiment, one or more processors 1002 include an integrated memory controller 1016 and a platform controller hub 1030. In at least one embodiment, the memory controller 1016 facilitates communication between memory devices and other components of the processing system 1000, while the platform controller hub (“PCH”) 1030 provides connectivity to input / output (“I / O”) devices via a local I / O bus.

[0308] In at least one embodiment, memory device 1020 may be a dynamic random access memory (“DRAM”) device, a static random access memory (“SRAM”) device, a flash memory device, a phase-change memory device, or other storage device with suitable performance for use as processor memory. In at least one embodiment, memory device 1020 may operate as system memory of processing system 1000 to store data 1022 and instructions 1021 for use by one or more processors 1002 when executing applications or processes. In at least one embodiment, memory controller 1016 is also coupled to an optional external graphics processor 1012, which may communicate with one or more graphics processors 1008 of processor 1002 to perform graphics and media operations. In at least one embodiment, display device 1011 may be connected to one or more processors 1002. In at least one embodiment, display device 1011 may include one or more internal display devices (such as mobile electronic devices or portable computer devices) or external display devices attached via a display interface (e.g., DisplayPort). In at least one embodiment, the display device 1011 may include a head-mounted display (“HMD”), such as a stereoscopic display device for virtual reality (“VR”) applications or augmented reality (“AR”) applications.

[0309] In at least one embodiment, the platform controller hub 1030 enables peripheral devices to connect to the memory device 1020 and the processor 1002 via a high-speed I / O bus. In at least one embodiment, the I / O peripheral devices include, but are not limited to, an audio controller 1046, a network controller 1034, a firmware interface 1028, a wireless transceiver 1026, a touch sensor 1025, and a data storage device 1024 (e.g., a hard disk drive, flash memory, etc.). In at least one embodiment, the data storage device 1024 may be connected via a storage interface (e.g., SATA) or via a peripheral bus (such as PCI or PCIe). In at least one embodiment, the touch sensor 1025 may include a touchscreen sensor, a pressure sensor, or a fingerprint sensor. In at least one embodiment, the wireless transceiver 1026 may be a Wi-Fi transceiver, a Bluetooth transceiver, or a mobile network transceiver (such as a 3G, 4G, or LTE transceiver). In at least one embodiment, the firmware interface 1028 enables communication with system firmware and may be, for example, a Unified Extensible Firmware Interface (“UEFI”). In at least one embodiment, network controller 1034 may enable network connectivity to a wired network. In at least one embodiment, a high-performance network controller (not shown) is coupled to interface bus 1010. In at least one embodiment, audio controller 1046 is a multi-channel high-definition audio controller. In at least one embodiment, processing system 1000 includes an optional legacy I / O controller 1040 for coupling legacy (e.g., Personal System 2 (“PS / 2”) devices to processing system 1000. In at least one embodiment, platform controller hub 1030 may also be connected to one or more Universal Serial Bus (“USB”) controllers 1042, connecting input devices (such as a keyboard and mouse combination 1043), camera 1044, or other USB input devices.

[0310] In at least one embodiment, instances of the memory controller 1016 and platform controller hub 1030 may be integrated into a discrete external graphics processor (such as external graphics processor 1012). In at least one embodiment, the platform controller hub 1030 and / or the memory controller 1016 may be external to one or more processors 1002. For example, in at least one embodiment, the processing system 1000 may include an external memory controller 1016 and a platform controller hub 1030, which may be configured to communicate with a memory controller hub and a peripheral controller hub within a system chipset of one or more processors 1002.

[0311] In at least one embodiment, the processing system 1000 can be used to implement system 700 (see...). Figure 7A ) and / or System 800 (see Figure 8In at least one embodiment, the first group 712 and / or the second group 714 may include one or more of one or more processors 1002, one or more of one or more processor cores 1007, and / or one or more of one or more graphics processors 1008. In at least one embodiment, the interface bus 1010 may be used to implement the switching circuit 710 (see...). Figure 7A ), one or more buses 728 and / or signal conductors 734 and 740. In at least one embodiment, utilizing Figure 10 The system described herein is one or more systems that implement one or more systems and / or processes, such as those described in conjunction with Figures 1-8.

[0312] Figure 11 The figure illustrates a computer system 1100 according to at least one embodiment. In at least one embodiment, the computer system 1100 may be a system having interconnect devices and components, a System-on-a-Chip (SoC), or a combination thereof. In at least one embodiment, the computer system 1100 is formed by a processor 1102, which may include an execution unit for executing instructions. In at least one embodiment, the computer system 1100 may include, but is not limited to, a component, such as the processor 1102, to employ an execution unit including logic to execute algorithms for processing data. In at least one embodiment, the computer system 1100 may include a processor, such as the PENTIUM® processor family, Xeon™, Itanium®, XScale™, and / or StrongArm™, Intel® Core™, or Intel® Nervana™ microprocessors, which are available from Intel Corporation, Santa Clara, California, but other systems (including PCs, engineering workstations, set-top boxes, etc.) with other microprocessors may also be used. In at least one embodiment, the computer system 1100 may execute a version of the Windows operating system available from Microsoft Corporation, Redmond, Washington, but other operating systems (such as UNIX and Linux), embedded software, and / or a graphical user interface may also be used.

[0313] In at least one embodiment, the computer system 1100 can be used in other devices, such as handheld devices and embedded applications. Some examples of handheld devices include cellular phones, Internet Protocol devices, digital cameras, personal digital assistants (“PDAs”), and handheld PCs. In at least one embodiment, the embedded application can include a microcontroller, a digital signal processor (DSP), a SoC, a network computer (“Net PC”), a set-top box, a network hub, a wide area network (“WAN”) switch, or any other system capable of executing one or more instructions.

[0314] In at least one embodiment, computer system 1100 may include, but is not limited to, processor 1102, which may include, but is not limited to, one or more execution units 1108, which may be configured to execute Unified Computing Device Architecture (“CUDA”) programs (CUDA® is developed by NVIDIA Corporation of Santa Clara, California). In at least one embodiment, a CUDA program is at least a portion of a software application written in the CUDA programming language. In at least one embodiment, computer system 1100 is a single-processor desktop or server system. In at least one embodiment, computer system 1100 may be a multiprocessor system. In at least one embodiment, processor 1102 may include, but is not limited to, a CISC microprocessor, a RISC microprocessor, a VLIW microprocessor, a processor implementing a combination of instruction sets, or any other processor device (such as a digital signal processor). In at least one embodiment, processor 1102 may be coupled to processor bus 1110, which may transmit data signals between processor 1102 and other components in computer system 1100.

[0315] In at least one embodiment, processor 1102 may include, but is not limited to, a level-one (“L1”) internal cache memory (“cache”) 1104. In at least one embodiment, processor 1102 may have a single internal cache or multiple levels of internal cache. In at least one embodiment, the cache memory may reside external to processor 1102. In at least one embodiment, processor 1102 may also include a combination of internal and external caches. In at least one embodiment, register file 1106 may store different types of data in various registers, including but not limited to integer registers, floating-point registers, status registers, and instruction pointer registers.

[0316] In at least one embodiment, execution unit 1108 (including, but not limited to, logic for performing integer and floating-point operations) also resides in processor 1102. Processor 1102 may also include a microcode (“ucode”) read-only memory (“ROM”) storing the microcode of certain macro instructions. In at least one embodiment, execution unit 1108 may include logic for processing packaged instruction set 1109. In at least one embodiment, by including packaged instruction set 1109 and associated circuitry for executing instructions in the instruction set of general-purpose processor 1102, packaged data in general-purpose processor 1102 can be used to perform operations used by numerous multimedia applications. In at least one embodiment, by using the full width of the processor data bus to perform operations on packaged data, the execution of numerous multimedia applications can be accelerated and more efficiently, eliminating the need to transfer smaller data units on the processor data bus to perform one or more operations on one data unit at a time.

[0317] In at least one embodiment, the execution unit 1108 may also be used in a microcontroller, embedded processor, graphics device, DSP, and other types of logic circuitry. In at least one embodiment, the computer system 1100 may include, but is not limited to, memory 1120. In at least one embodiment, memory 1120 may be implemented as a DRAM device, SRAM device, flash memory device, or other memory device. Memory 1120 may store one or more instructions 1119 and / or data 1121 represented by data signals executable by processor 1102.

[0318] In at least one embodiment, the system logic chip may be coupled to processor bus 1110 and memory 1120. In at least one embodiment, the system logic chip may include, but is not limited to, a memory controller hub (“MCH”) 1116, and processor 1102 may communicate with MCH 1116 via processor bus 1110. In at least one embodiment, MCH 1116 may provide a high-bandwidth memory path 1118 to memory 1120 for instruction and data storage, as well as for storage of graphics commands, data, and textures. In at least one embodiment, MCH 1116 may direct data signals between processor 1102, memory 1120, and other components in computer system 1100, and bridge data signals between processor bus 1110, memory 1120, and system I / O 1122. In at least one embodiment, the system logic chip may provide a graphics port for coupling to a graphics controller. In at least one embodiment, MCH 1116 can be coupled to memory 1120 via high-bandwidth memory path 1118, and graphics / video card 1112 can be coupled to MCH 1116 via Accelerated Graphics Port (“AGP”) interconnect 1114.

[0319] In at least one embodiment, computer system 1100 may use system I / O 1122, which serves as a proprietary hub interface bus, to couple MCH 1116 to I / O controller hub (“ICH”) 1130. In at least one embodiment, ICH 1130 may provide direct connectivity to some I / O devices via a local I / O bus. In at least one embodiment, the local I / O bus may include, but is not limited to, a high-speed I / O bus for connecting peripheral devices to memory 1120, chipset, and processor 1102. Examples may include, but are not limited to, an audio controller 1129, a firmware hub (“flash BIOS”) 1128, a wireless transceiver 1126, data storage 1124, a conventional I / O controller 1123 including a user input interface 1125 and a keyboard interface, a serial expansion port 1127 (such as USB), and a network controller 1134. Data storage 1124 may include a hard disk drive, floppy disk drive, CD-ROM device, flash memory device, or other mass storage device.

[0320] In at least one embodiment, Figure 11 The diagram illustrates a system including interconnected hardware devices or "chips". In at least one embodiment, Figure 11 An example SoC can be illustrated in the diagram. In at least one embodiment, in Figure 11 The devices illustrated herein may be interconnected using proprietary interconnects, standardized interconnects (e.g., PCIe), or a combination thereof. In at least one embodiment, one or more components of system 1100 are interconnected using a compute fast link (“CXL”) interconnect.

[0321] In at least one embodiment, computer system 1100 can be used to implement system 700 (see Figure 7A ) and / or System 800 (see Figure 8 In at least one embodiment, the first group 712 and / or the second group 714 may include a processor 1102 and / or a graphics / video card 1112. In at least one embodiment, the processor bus 1110 may be used to implement a switching circuit 710 (see...). Figure 7A ), one or more buses 728 and / or signal conductors 734 and 740. In at least one embodiment, utilizing Figure 11 The system described herein is one or more systems that implement one or more systems and / or processes, such as those described in conjunction with Figures 1-8.

[0322] Figure 12The figure illustrates a system 1200 according to at least one embodiment. In at least one embodiment, system 1200 is an electronic device utilizing processor 1210. In at least one embodiment, system 1200 may be, for example, but not limited to, a laptop computer, tower server, rack server, blade server, edge device communicatively coupled to one or more enterprise on-premises or cloud service providers, portable computer, desktop computer, tablet computer, mobile device, telephone, embedded computer, or any other suitable electronic device.

[0323] In at least one embodiment, system 1200 may include, but is not limited to, processor 1210 communicatively coupled to any suitable number or type of components, peripherals, modules, or devices. In at least one embodiment, processor 1210 is coupled using a bus or interface such as an I2C bus, a system management bus (“SMBus”), a low pin count (“LPC”) bus, a serial peripheral interface (“SPI”), a high-definition audio (“HDA”) bus, a serial advanced technology accessory (“SATA”) bus, a USB (versions 1, 2, 3) bus, or a universal asynchronous receiver / transmitter (“UART”) bus. In at least one embodiment, Figure 12 The diagram illustrates a system including interconnected hardware devices or "chips," in at least one embodiment, Figure 12 An example SoC can be illustrated. In at least one embodiment, in Figure 12 The devices illustrated herein can be interconnected via proprietary interconnects, standardized interconnects (such as PCIe), or some combination thereof. In at least one embodiment, Figure 12 One or more components are interconnected using CXL interconnects.

[0324] In at least one embodiment, Figure 12 This may include a display 1224, a touchscreen 1225, a touchpad 1230, a near-field communication unit (“NFC”) 1245, a sensor hub 1240, a thermal sensor 1246, a fast chipset (“EC”) 1235, a trusted platform module (“TPM”) 1238, a BIOS / firmware / flash (“BIOS, FW flash”) 1222, a DSP 1260, a solid-state drive (“SSD”) or hard disk drive (“HDD”) 1220, a wireless local area network unit (“WLAN”) 1250, a Bluetooth unit 1252, a wireless wide area network unit (“WWAN”) 1256, a global positioning system (“GPS”) 1255, a camera such as a USB 3.0 camera (“USB 3.0 camera”) 1254, or a low-power double data rate (“LPDDR”) memory unit (“LPDDR3”) 1215 implemented in, for example, the LPDDR3 standard. These components may be implemented in any suitable manner.

[0325] In at least one embodiment, other components may be communicatively coupled to processor 1210 via the components discussed above. In at least one embodiment, accelerometer 1241, ambient light sensor (“ALS”) 1242, compass 1243, and gyroscope 1244 may be communicatively coupled to sensor hub 1240. In at least one embodiment, thermal sensor 1239, fan 1237, keyboard 1236, and touchpad 1230 may be communicatively coupled to EC 1235. In at least one embodiment, speaker 1263, earphone 1264, and microphone (“mic”) 1265 may be communicatively coupled to audio unit (“audio codec and Class D amplifier”) 1262, which in turn may be communicatively coupled to DSP 1260. In at least one embodiment, audio unit 1262 may include, for example, but not limited to, an audio encoder / decoder (“codec”) and a Class D amplifier. In at least one embodiment, SIM card (“SIM”) 1257 may be communicatively coupled to WWAN unit 1256. In at least one embodiment, components such as WLAN unit 1250, Bluetooth unit 1252, and WWAN unit 1256 may be implemented in a next-generation form factor (“NGFF”).

[0326] In at least one embodiment, system 1200 can be used to implement system 700 (see...). Figure 7A ) and / or System 800 (see Figure 8 In at least one embodiment, processor 722 (see...) Figure 7A This can be implemented using processor 1210. In at least one embodiment, utilizing... Figure 12 The system described herein is one or more systems that implement one or more systems and / or processes, such as those described in conjunction with Figures 1-8.

[0327] Figure 13The figure illustrates an example integrated circuit 1300 according to at least one embodiment. In at least one embodiment, the example integrated circuit 1300 is a SoC (System-on-a-Chip) that can be manufactured using one or more IP cores. In at least one embodiment, the integrated circuit 1300 includes one or more application processors 1305 (e.g., CPU, DPU), at least one graphics processor 1310, and may further include an image processor 1315 and / or a video processor 1320, any of which may be a modular IP core. In at least one embodiment, the integrated circuit 1300 includes peripheral or bus logic, including a USB controller 1325, a UART controller 1330, an SPI / SDIO controller 1335, and an I2S / I2C controller 1340. In at least one embodiment, the integrated circuit 1300 may include a display device 1345 coupled to one or more of a High Definition Multimedia Interface (“HDMI”) controller 1350 and a Mobile Industrial Processor Interface (“MIPI”) display interface 1355. In at least one embodiment, memory may be provided by a flash memory subsystem 1360 including flash memory and a flash memory controller. In at least one embodiment, a memory interface for accessing an SDRAM or SRAM memory device may be provided via a memory controller 1365. In at least one embodiment, some integrated circuits also include an embedded security engine 1370.

[0328] In at least one embodiment, integrated circuit 1300 can be used to implement system 700 (see...). Figure 7A ) and / or System 800 (see Figure 8 In at least one embodiment, the first group 712 and / or the second group 714 may include one or more of one or more application processors 1305, one or more of one or more graphics processors 1310, image processors 1315, and / or video processors 1320. (See reference...) Figure 7A In at least one embodiment, integrated circuit 1300 (see...) Figure 13 The peripheral or bus logic of ) can be used to implement the switching circuit 710 (see Figure 7A ), one or more buses 728 and / or signal conductors 734 and 740.

[0329] Figure 14The figure illustrates a computing system 1400 according to at least one embodiment. In at least one embodiment, the computing system 1400 includes a processing subsystem 1401 having one or more processors 1402 and a system memory 1404 communicating via an interconnect path, which may include a memory hub 1405. In at least one embodiment, the memory hub 1405 may be a separate component within a chipset assembly or integrated within one or more processors 1402. In at least one embodiment, the memory hub 1405 is coupled to an I / O subsystem 1411 via a communication link 1406. In at least one embodiment, the I / O subsystem 1411 includes an I / O hub 1407 that enables the computing system 1400 to receive input from one or more input devices 1408. In at least one embodiment, the I / O hub 1407 enables a display controller, which may be included in one or more processors 1402, to provide output to one or more display devices 1410A. In at least one embodiment, one or more display devices 1410A coupled to the I / O hub 1407 may include local, internal, or embedded display devices.

[0330] In at least one embodiment, the processing subsystem 1401 includes one or more parallel processors 1412 coupled to a memory hub 1405 via a bus or other communication link 1413. In at least one embodiment, the communication link 1413 may be one of any number of standards-based communication link technologies or protocols, such as, but not limited to, PCIe, or may be a vendor-specific communication interface or communication architecture. In at least one embodiment, the one or more parallel processors 1412 form a computation-centric parallel or vector processing system that may include a large number of processing cores and / or processing clusters, such as many integrated core processors. In at least one embodiment, the one or more parallel processors 1412 form a graphics processing subsystem that can output pixels to one of one or more display devices 1410A coupled via an I / O hub 1407. In at least one embodiment, the one or more parallel processors 1412 may also include a display controller and a display interface (not shown) for direct connection to one or more display devices 1410B.

[0331] In at least one embodiment, system storage unit 1414 may be connected to I / O hub 1407 to provide a storage mechanism for computing system 1400. In at least one embodiment, I / O switch 1416 may be used to provide an interface mechanism to enable connectivity between I / O hub 1407 and other components, such as network adapter 1418 and / or wireless network adapter 1419 that may be integrated into the platform, and various other devices that may be added via one or more accessory devices 1420. In at least one embodiment, network adapter 1418 may be an Ethernet adapter or another wired network adapter. In at least one embodiment, wireless network adapter 1419 may include one or more of Wi-Fi, Bluetooth, NFC, or other network devices including one or more wireless radios.

[0332] In at least one embodiment, the computing system 1400 may include other components not explicitly shown, including USB or other port connectors, optical storage drives, video capture devices, etc., which may also be connected to the I / O hub 1407. In at least one embodiment, in Figure 14 The communication paths that interconnect various components can be implemented using any suitable protocol, such as PCI-based protocols (e.g., PCIe), or other bus or point-to-point communication interfaces and / or protocols (e.g., NVLink™ high-speed interconnect), or interconnect protocols.

[0333] In at least one embodiment, one or more parallel processors 1412 include circuitry optimized for graphics and video processing (including, for example, video output circuitry) and constitute a graphics processing unit (“GPU”). In at least one embodiment, one or more parallel processors 1412 include circuitry optimized for general-purpose processing. In at least one embodiment, components of the computing system 1400 may be integrated with one or more other system elements on a single integrated circuit. For example, in at least one embodiment, one or more parallel processors 1412, memory hub 1405, processor 1402, and I / O hub 1407 may be integrated into a SoC integrated circuit. In at least one embodiment, components of the computing system 1400 may be integrated into a single package to form a system-in-package (“SIP”) configuration. In at least one embodiment, at least a portion of the components of the computing system 1400 may be integrated into a multi-chip module (“MCM”) that may interconnect with other multi-chip modules to form a modular computing system. In at least one embodiment, the I / O subsystem 1411 and display device 1410B are omitted from the computing system 1400.

[0334] In at least one embodiment, computing system 1400 may be used to implement system 700 (see FIG. 7) and / or system 800 (see FIG. 8). Figure 8 In at least one embodiment, the first group 712 and / or the second group 714 may include one or more of one or more processors 1402 and / or one or more of one or more parallel processors 1412. In at least one embodiment, the communication link 1413 may be used to implement the switching circuit 710 (see FIG. 7), one or more buses 728 and / or signal conductors 734 and 740.

[0335] It should be noted that while the example embodiments described herein may relate to the CUDA programming model, the techniques described herein can be used with any suitable programming model, such as HIP, oneAPI (e.g., using oneAPI-based programming to perform or implement the methods disclosed herein) and / or variations thereof.

[0336] In at least one embodiment, one or more components of the system and / or processor disclosed above may communicate with one or more CPUs, ASICs, GPUs, FPGAs, or other hardware, circuit systems, or integrated circuit components, including, for example, amplifiers or upsamplers for amplifying images, image fusion units or image fusion unit components for fusing, mixing, or adding images together, samplers for sampling images (e.g., as part of a DSP), neural network circuitry configured to perform amplifiers to amplify images (e.g., from low-resolution images to high-resolution images), or other hardware for modifying or generating images, frames, or videos to adjust their resolution, size, or pixels; one or more components of the system and / or processor disclosed above may use the components described in this disclosure to perform methods, operations, or instructions for generating or modifying images.

[0337] It should be understood that any feature described herein may be claimed in combination with one or more other features as described herein, regardless of whether those features are derived from the same embodiment described.

[0338] Specific details are set forth in the description to provide a thorough understanding of the embodiments. However, those skilled in the art will understand that the embodiments can be practiced without these specific details. In other instances, well-known circuits, processes, algorithms, structures, and techniques may be shown without unnecessary detail so as not to obscure the embodiments.

[0339] While illustrative embodiments of the present disclosure have been described in detail herein, it should be understood that the concepts of the invention may be embodied and used in different ways, and the appended claims are intended to be construed as including these variations, except where limited by the prior art. It should be understood that any feature described herein may be combined with any other one or more features described herein to form a claim, regardless of whether these features are derived from the same embodiment.

Claims

1. A network interface device, the network interface device comprising: An optical interface is used to transmit data via an optical communication medium. A wireless interface used to transmit data via a wireless communication medium; and One or more circuits are configured to selectively route data from a host device to at least one of the optical interface or the wireless interface in response to a threshold condition.

2. The network interface device according to claim 1, wherein, The one or more circuits are also used for at least one of the following: Switching data transmission from the first communication medium to the second communication medium; or Two communication media can be used concurrently.

3. The network interface device according to claim 1, wherein, The one or more circuits are also configured to route data via both the optical interface and the wireless interface in response to the threshold condition.

4. The network interface device according to claim 1, wherein, The threshold condition is based on at least one of the following: A measured communication link quality metric associated with at least one of the optical communication medium or the wireless communication medium; Measurement of bandwidth utilization associated with at least one of the optical communication medium or the wireless communication medium; The detected bit error rate associated with at least one of the optical communication medium or the wireless communication medium; or The signal-to-noise ratio associated with at least one of the optical communication medium or the wireless communication medium.

5. The network interface device according to claim 1, wherein, The one or more circuits are also used for: Receive instructions regarding workload requirements; and The indication of the workload demand is compared with the threshold condition.

6. The network interface device according to claim 5, wherein, The one or more circuits are also used for: High-priority workload traffic is routed through the optical interface; and Lower-priority workload traffic is routed through the wireless interface.

7. The network interface device according to claim 1, wherein, The one or more circuits include a local memory storing AI-based decision logic, and the threshold condition is determined by the AI-based decision logic in response to usage metrics collected from the optical interface and the wireless interface.

8. The network interface device according to claim 1, wherein, The threshold condition is triggered in response to receiving a switching command from an external orchestration controller in the data center network.

9. The network interface device according to claim 1, further comprising a coupling interface for coupling the network interface device to the host device, wherein, The coupling interface is also used to communicate with the host device using at least one of an optical path or an electrical path.

10. The network interface device according to claim 1, wherein, The wireless interface includes at least one of the following: A photonic mixer stage is used to generate optical signals; or The radio frequency mixer stage is configured to generate or receive millimeter-wave or terahertz signals.

11. The network interface device according to claim 1, further comprising: A photon signal generator is used to generate modulated optical signals; and A branch module is used to direct the modulated optical signal to at least one of the optical interface or the wireless interface.

12. The network interface device according to claim 11, further comprising an optical output that couples the modulated optical signal to an optical fiber.

13. The network interface device according to claim 12, wherein, The wireless interface includes: Photonic mixer stage, used to generate millimeter-wave or terahertz signals; and An antenna coupled to the photonic mixer stage.

14. The network interface device according to claim 1, wherein, The one or more circuits are used to adaptively adjust the operation of the network interface device based on at least one of real-time network conditions or workload requirements, the operation of adjusting the network interface device including at least one of: modifying data routing logic, selecting between communication media, or reallocating bandwidth on the optical interface and the wireless interface.

15. A network interface device, the network interface device comprising: An optical interface is used to receive modulated optical signals; An optical processing circuit is used to demodulate the modulated optical signal; Wireless interface, used to receive wireless signals; A wireless processing circuit for decoding the wireless signal; and One or more output circuits coupled to the optical processing circuit and the wireless processing circuit, the one or more output circuits being used to provide demodulated data to the host system.

16. The network interface device of claim 15, further comprising a controller communicatively coupled to the optical processing circuit and the wireless processing circuit, the controller being configured to: Monitor at least one link quality metric associated with the optical interface and the wireless interface; and Selectively enable or disable reception via at least one of the optical interface or wireless interface based on threshold conditions.

17. The network interface device according to claim 16, wherein, The controller is also configured to receive data concurrently through the optical interface and the wireless interface in response to detecting that at least one of the total bandwidth requirement or link utilization meets or exceeds the threshold condition.

18. The network interface device according to claim 16, wherein, The controller includes AI-based decision logic for modifying the threshold conditions based on the at least one link quality metric.

19. A system comprising: Host device, used to generate outgoing data traffic; A network interface device, communicatively coupled to the host device, the network interface device comprising: An optical interface is used to transmit data via an optical communication medium. A wireless interface is used to transmit data via a wireless communication medium, and A controller is configured to selectively route the outgoing data traffic between the optical interface and the wireless interface based on threshold conditions.

20. The system according to claim 19, wherein, The network interface device also includes: Optical processing circuitry for demodulating incoming optical signals from the optical communication medium; and A wireless processing circuit is used to decode incoming wireless signals from the wireless communication medium, and wherein the controller is further used to route incoming data to the host device.

21. The system according to claim 20, wherein, The controller includes an artificial intelligence module for: Monitor link performance metrics from the optical interface and the wireless interface; and The threshold conditions are dynamically adjusted in response to changes in the link performance metric.

22. The system according to claim 19, wherein, The host device includes a workload manager that assigns priority indicators to outgoing data, and the controller is configured to route higher priority traffic via the optical interface and lower priority traffic via the wireless interface.

23. A hybrid network switch, the hybrid network switch comprising: Multiple wired ports are used for sending and receiving data via physical cabling; Multiple wireless ports are used to send and receive data via wireless communication media; and Switching hardware coupled to the plurality of wired ports and the plurality of wireless ports, the switching hardware being used to route data from an ingress port to a selected egress port based on routing or switching logic.

Images