Photoelectric collaborative data high-speed exchange system and communication equipment

Abstract

The invention provides a photoelectric collaborative data high-speed switching system. The photoelectric collaborative data high-speed switching system comprises an electric processing unit, an optical switching unit and an optical switching controller, the electric processing unit receives a data frame, splits the data frame into a plurality of cell slices, and allocates a cache address to each cell slice; the optical switching controller receives the request sending message, inquires wavelength scheduling configuration information according to destination address information in the request sending message, and generates a wavelength scheduling control instruction; the optical switching unit receives the wavelength scheduling control instruction, controls the conduction state of the controllable optical switch array according to the wavelength scheduling control instruction, and establishes an optical channel from a source end to a destination end by using the wavelength routing characteristic of the arrayed waveguide grating router; and the electric processing unit is also integrated with a remote direct memory access unit which outputs the cache address and the data length of each cell slice after receiving the sending permission message, so that the target-end electric processing unit directly reads the cell slices from the source-end cache and recombines the cell slices into data frames.

Description

Technical Field

[0001] This disclosure relates to the field of communication technology, specifically to a photoelectric collaborative high-speed data exchange system and communication equipment. Background Technology

[0002] In data centers and high-performance computing scenarios, traditional electrical switching architectures face bottlenecks such as limited bandwidth, high latency, and high power consumption. To improve switching performance, optoelectronic converged switching technology has gradually become a research hotspot.

[0003] Existing solutions, such as Chinese patent application CN117499289A (publication date: February 6, 2024), propose a data center interconnection method based on optical circuit switching. This method achieves low-latency transmission through pre-established optical channels and employs a request-to-send / allow-to-send handshake mechanism to coordinate communication. However, this solution suffers from the following technical defects that are seriously inconsistent with practical application requirements: (1) Lack of fine-grained data processing mechanism: Its electrical processing unit directly initiates optical switching requests in units of complete data frames without dividing the data frames into smaller transmission units, resulting in excessively long single optical channel occupancy time, which cannot adapt to bursty, small-grained service flows and limits the system's concurrency capability. (2) Rigid optical resource scheduling: Once the optical switching unit is established, it monopolizes a specific wavelength path. Even if the actual data volume is much smaller than the channel capacity, it cannot dynamically release or reuse the wavelength, resulting in a serious waste of wavelength routing resources such as arrayed waveguide grating routers (AWGR). (3) No integrated zero-copy data path: Its remote direct memory access mechanism is missing or not coordinated with the optical switching process. The destination still needs to rely on the CPU to participate in data transfer and reassembly, and cannot directly expose and read the source cache address, which restricts the end-to-end throughput performance. (4) The control plane lacks online update capability: the wavelength scheduling configuration information in the optical switching controller is managed statically in a single table. When updating, the existing services need to be interrupted, which cannot support batch modification and seamless switching, making it difficult to adapt to the needs of dynamic topology and traffic changes in large-scale networks.

[0004] The aforementioned shortcomings make it difficult for existing optoelectronic switching systems to balance resource efficiency, flexibility, and performance in high-density cross-connect, broadcast pass-through, and low-latency RDMA application scenarios. There is an urgent need for a new optoelectronic collaborative switching architecture that can simultaneously achieve data slicing-driven, wavelength-on-demand scheduling, direct address output with zero copy, and hot configuration updates. Summary of the Invention

[0005] This disclosure provides a photoelectric collaborative high-speed data exchange system and communication equipment.

[0006] According to one aspect of this disclosure, a high-speed optoelectronic collaborative data exchange system is provided, comprising an electrical processing unit, an optical switching unit, and an optical switching controller; wherein, the electrical processing unit is configured to: receive a data frame, divide the data frame into multiple cell slices, and allocate a buffer address to each cell slice; the optical switching controller is configured to: receive a request to send a message, query wavelength scheduling configuration information based on the destination address information in the request to send the message, and generate a wavelength scheduling control command; the optical switching unit includes a controllable optical switch array and an arrayed waveguide grating router, and is configured to: receive the wavelength scheduling control command, control the conduction state of the controllable optical switch array according to the wavelength scheduling control command, and establish an optical channel from the source end to the destination end using the wavelength routing characteristics of the arrayed waveguide grating router; the electrical processing unit also integrates a remote direct memory access unit, which is configured to: after receiving a permission to send a message, output the buffer address and data length of each cell slice, so that the destination end electrical processing unit can directly read the cell slice from the source end buffer based on the buffer address and data length and reassemble it into a data frame.

[0007] According to the technical solution of this embodiment, by dividing the data frame into cell slices, the single optical channel occupancy time can be shortened and the concurrency capability can be improved; by scheduling the optical channel on demand at the wavelength level, resource monopoly can be avoided and wavelength utilization can be improved; by outputting the cache address of the remote direct memory access unit for direct reading by the destination end, CPU participation can be eliminated and zero-copy low-latency transmission can be achieved.

[0008] According to at least one embodiment of the system of this disclosure, the electrical processing unit includes an electrical switching function module and a control agent; the control agent is configured to: process the data frame, and when it is determined that the data frame is a broadcast frame, send a broadcast request transmission message to the optical switching controller.

[0009] According to the technical solution of this embodiment, by controlling the agent to identify broadcast frames and trigger broadcast requests, it is possible to avoid copying multiple complete frames in the electrical domain, thereby reducing cache overhead and processing latency.

[0010] According to at least one embodiment of the system of this disclosure, the remote direct memory access unit is configured to output the cache address and data length of each cell slice via a high-speed serial bus.

[0011] According to the technical solution of this embodiment, the cache address and data length of the cell slice are output through a high-speed serial bus, which can realize low-latency, high-bandwidth address notification and support zero-copy direct reading at the destination.

[0012] According to at least one embodiment of the system of this disclosure, the optical switching unit is connected to the electrical processing unit via a photoelectric conversion interface; the photoelectric conversion interface is configured to: convert electrical signals into multiple optical signals and transmit them to the optical switching unit via a transmit optical channel, and receive multiple optical signals via a receive optical channel and convert them into electrical signals; the transmit optical channel and the receive optical channel are respectively used for unidirectional optical transmission of data.

[0013] According to the technical solution of this embodiment, by connecting the optical switching unit and the electrical processing unit through a photoelectric conversion interface with transmit and receive optical channels, efficient bidirectional conversion between electrical signals and multiple optical signals can be achieved, thereby supporting fine-grained optical switching driven by data slicing, on-demand dynamic scheduling of wavelength resources, zero-copy pass-through transmission, and online hot update of control configuration, effectively breaking through the bottlenecks of traditional electrical switching architecture in terms of bandwidth, latency, and power consumption.

[0014] According to at least one embodiment of the system of this disclosure, the controllable optical switch array is implemented using a semiconductor optical amplifier; the semiconductor optical amplifier is configured to control its on or off state by injecting current and has a nanosecond-level switching rate; the optical switching unit further includes a high-speed high-current drive circuit unit, a temperature control circuit unit, and a gain control circuit unit.

[0015] According to the technical solution of this embodiment, by constructing a controllable optical switch array using a semiconductor optical amplifier and integrating high-speed high-current drive, temperature control and gain control circuits, nanosecond-level optical switch response and stable gain output can be achieved, effectively supporting the rapid optical path establishment and release of fine-grained data slices, improving wavelength resource utilization efficiency, and reducing switching delay and power consumption.

[0016] According to at least one embodiment of the system of this disclosure, the optical switching controller stores two sets of wavelength scheduling configuration information, one set being active and the other set being inactive; the optical switching controller is configured to: perform batch updates on the wavelength scheduling configuration information in the inactive state, and switch the wavelength scheduling configuration information to the active state after the update is completed.

[0017] According to the technical solution of this embodiment, by maintaining two sets of wavelength scheduling configuration information in the optical switching controller and supporting batch updates and seamless switching of inactive configurations, online hot updates of the control plane can be achieved, avoiding service interruption and meeting the needs of flexible and highly available optical resource scheduling in large-scale dynamic networks.

[0018] According to at least one embodiment of the system of this disclosure, the optical switching controller is configured to: upon receiving the broadcast request to send message, return a broadcast permission to send message to the electrical processing unit; the optical switching unit is configured to: upon receiving the broadcast permission to send message, perform broadcast pass-through through all optical channels.

[0019] According to the technical solution of this embodiment, by responding to broadcast requests and triggering full-channel broadcast pass-through through the optical switching controller, low-latency and high-efficiency multicast / broadcast service support can be achieved, avoiding bandwidth waste and processing bottlenecks caused by hop-by-hop replication in traditional electrical switching.

[0020] According to at least one embodiment of the system disclosed herein, a plurality of said optical switching units are interconnected according to the Dragonfly+ topology rule.

[0021] According to the technical solution of this embodiment, by interconnecting multiple optical switching units in a Dragonfly+ topology, the network diameter and congestion probability can be significantly reduced, the bandwidth scalability and fault tolerance under high-density interconnection can be improved, and the latency and power consumption bottlenecks of traditional electrical switching architecture in large-scale deployment can be effectively alleviated.

[0022] According to at least one embodiment of the present disclosure, the system is deployed in a cross-connection scenario of communication equipment and includes multiple service boards and at least one cross-connect board; the optical switching unit and the optical switching controller are deployed in the cross-connect board.

[0023] According to the technical solution of this embodiment, by integrating optical switching units and optical switching controllers in the cross-connect board, high-bandwidth, low-latency optoelectronic collaborative cross-connection between service boards can be realized, breaking through the limitations of traditional electrical cross-connection in terms of power consumption and scalability, and meeting the needs of communication equipment for high-density, high-efficiency interconnection.

[0024] According to another aspect of this disclosure, a communication device is provided, including the optoelectronic cooperative high-speed data exchange system as described above. Attached Figure Description The accompanying drawings illustrate exemplary embodiments of the present disclosure and, together with the description thereof, serve to explain the principles of the present disclosure. These drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification.

[0025] Figure 1 This is a diagram of an optoelectronic collaborative high-speed data exchange system architecture according to one embodiment of the present disclosure.

[0026] Figure 2 This is a schematic diagram of a photoelectric conversion entity according to one embodiment of the present disclosure. Figure 3This is a diagram of a programmable SOA array module according to one embodiment of the present disclosure.

[0027] Figure 4 This is a block diagram of the interconnection between a PCIe3-based switching controller and various switching agents according to one embodiment of this disclosure.

[0028] Figure 5 This is an interconnection diagram of a PCIe3-based switching controller and SOA switch arrays of each OWS according to one embodiment of the present disclosure.

[0029] Figure 6 This is a diagram of a frame transmission process (source end) according to one embodiment of the present disclosure.

[0030] Figure 7 This is a diagram of a frame reception process (sink end) according to one embodiment of the present disclosure.

[0031] Figure 8 This is a diagram of an efficient update mechanism for the OWS optical channel mapping table according to one embodiment of the present disclosure.

[0032] Figure 9 This is a flowchart of the optoelectronic cooperative switching control surface according to one embodiment of the present disclosure.

[0033] Figure 10 This is a flowchart of the optoelectronic cooperative exchange data plane according to one embodiment of the present disclosure.

[0034] Figure 11 This is a silicon photonic integrated system-on-chip deployment diagram according to one embodiment of the present disclosure.

[0035] Figure 12 This is a deployment diagram of an optoelectronic cross-connection scheme for a communication device according to one embodiment of the present disclosure.

[0036] Figure 13 This is an implementation deployment diagram of a communication equipment optoelectronic cross-connection expansion scheme according to one embodiment of the present disclosure.

[0037] Figure 14 This is an implementation deployment diagram of a communication device optoelectronic cross-connect derating scheme according to one embodiment of this disclosure. Detailed Implementation

[0038] The present disclosure will now be described in further detail with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are for illustrative purposes only and are not intended to limit the scope of the disclosure. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the accompanying drawings.

[0039] It should be noted that, where there is no conflict, the embodiments and features described in this disclosure can be combined with each other. The technical solutions of this disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0040] In high-density cross-connect scenarios (such as X16 slots and communication equipment with a single slot capacity of 1.6T), existing optoelectronic switching systems struggle to balance low latency and high throughput. After careful research, the inventors discovered the fundamental reason lies in: (1) Scheduling optical channels in units of complete data frames results in microsecond-level channel establishment overhead being much greater than nanosecond-level actual transmission time, causing a large amount of wavelength resources to be wasted. (2) Broadcast services require multiple copies of the complete frame in the electrical domain, increasing the buffer burden and introducing additional latency; (3) The control plane uses a single static configuration table, which makes it impossible to update the scheduling strategy without interrupting the service.

[0041] To address this, the present disclosure proposes the following technical solution: In this solution, an optoelectronic collaborative high-speed data exchange system is used. The electrical processing unit segments data frames into cell slices and allocates buffer addresses. The optical switching controller queries the wavelength scheduling configuration based on the destination address and generates control commands. The optical switching unit then establishes a wavelength-level optical channel. Simultaneously, a remote direct memory access unit integrated into the electrical processing unit outputs the buffer address and length of the cell slices after the channel is established, allowing the destination end to directly read and reassemble them. This solution achieves fine-grained scheduling, zero-copy transmission, and hot configuration updates, significantly improving optical resource utilization and system throughput.

[0042] This technical solution achieves high throughput, low latency, and high resource efficiency optoelectronic converged switching without relying on specific optical devices, effectively meeting the performance requirements of high-density cross-connection in large-scale data centers.

[0043] To facilitate description and make the technical solutions of this disclosure easier to understand, the terminology of this disclosure will be explained before describing the technical solutions of this disclosure.

[0044] ONet module (Optical Network module) OWS module (Optical Wavelength Switching Unit) CPO-TR module Co-Packaged Optics Transceiver AWGR-TTunable Arrayed Waveguide Grating Router SOAS semiconductor optical amplifier RTSRequest To Send CTSClear To Send (Allow Sending) CTL(n) Control Signal PCIe3X4 Peripheral Component Interconnect Express Gen3×4 (High-speed serial bus standard for transmitting RTS / CTS and CTL signals between ONet and the controller) RDMARemote Direct Memory Access Figure 1 This disclosure illustrates an architecture diagram of a high-speed optoelectronic collaborative data switching system according to one embodiment, including an electrical processing unit, an optical switching unit, and an optical switching controller. The electrical processing unit is configured to: receive data frames, segment the data frames into multiple cell slices, and allocate buffer addresses to each cell slice. The optical switching controller is configured to: receive request transmission messages, query wavelength scheduling configuration information based on the destination address information in the request transmission message, and generate wavelength scheduling control instructions. The optical switching unit includes a controllable optical switch array and an arrayed waveguide grating router. It is configured to: receive the wavelength scheduling control command, and control the conduction state of the controllable optical switch array according to the wavelength scheduling control command, and establish an optical channel from the source end to the destination end by utilizing the wavelength routing characteristics of the array waveguide grating router; the electrical processing unit also integrates a remote direct memory access unit, which is configured to: after receiving a permission to send message, output the cache address and data length of each cell slice, so that the destination end electrical processing unit can directly read the cell slice from the source end cache based on the cache address and data length and reassemble it into a data frame.

[0045] The system will be explained in three parts: the optoelectronic collaborative high-speed data exchange device, the optoelectronic collaborative exchange control device, and the basic operation procedure of the system. The optoelectronic collaborative high-speed data switching device includes an electrical processing unit, an optical switching unit, and an optical switching controller. The optoelectronic collaborative switching device consists of two main parts: the ONet module (electrical processing unit) and the OWS module (containing the optical switching unit and optical switching controller). Figure 1As shown, the ONet module provides four 400Gbps Ethernet service interfaces. The ONet module's switching agent, as a slave device, is interconnected with the OWS module's switching controller master device via a PCIe3 X4 bus, serving as the underlying communication channel between the two. The ONet module's CPO-TR module is interconnected with the OWS module's SOA array via two 16-optical fibers, one for data transmission and the other for data reception.

[0046] The ONet module consists of four technical components: an electrical switching function module, a switching agent, an RDMA network interface card module, and a CPO-TR module. The functional definitions of each component are as follows: 1) Electrical switching function module: Implements traditional switching or NP functions, such as PHY / MAC / PP / TM, FLEXE slicing, OAM, QoS, etc., and can adopt existing functional entities in the industry.

[0047] 2) Switching Agent: This acts as an agent for the switching controller in each ONET module of the device system. Its main functions include: The data frame of port 5 of the A transceiver module is divided into 4 equal parts into cells, and the data of the cell slice is placed in the embedded buffer.

[0048] B is responsible for sending request commands (RTS) and send enable commands (CTS) to the switching controller. These commands are transmitted via the PCIe3 bus. C sends RDMA transmit / receive commands to the RDMA network card module via the PCIe5 bus. The commands include information such as the storage address of data.

[0049] 3) RDMA network card module: Four 800Gbps RDMA network cards, which are interconnected with the main device of the switching agent via PCIe5 ×32 respectively. After receiving instructions from the switching agent, they send and receive metadata according to the stored address.

[0050] 4) CPO-TR module: This is the photoelectric conversion functional entity of the ONET module.

[0051] The OWS module consists of four technical components: a switching controller, a programmable SOA switch array, an arrayed waveguide grating router (AWGR), and a demultiplexer (DeMux). The functions of each component are defined as follows: 1) Switching Controller: Responsible for the overall control of the optical channels for system data switching, specifically as follows: A is responsible for command processing and response distribution for each exchange agent.

[0052] B is responsible for the orchestration and control of the wavelength parallel queue output of the programmable SOA switch array, which is achieved through PCIe3 bus CTL commands.

[0053] 2) Programmable SOA Array: A accepts programmable signal control from the switching controller.

[0054] B. Reorder the 16 light waves as needed.

[0055] 3) Arrayed Waveguide Grating Router (AWGR-T): The input signals from different input ports have a cyclic shift characteristic at the output port: the wavelengths of light from different inputs going to the same output must be different, thus eliminating port contention in the routing and realizing non-blocking switching of optical packets.

[0056] 4) Demultiplexer (DeMux): Primarily a wavelength division multiplexer, it reorders the 16 wavelengths of any one of the 16 optical links into parallel sequences λ1…λ. 16 .

[0057] The core technologies of the optoelectronic collaborative high-speed data switching device (including optical switching controller and optical switching unit) are optoelectronic conversion technology and high-speed optical switching technology. The optoelectronic conversion technology involves network card modules and CPO-TR modules that realize the data transmission and reception functions of the optical switching data plane, while the high-speed optical switching technology involves programmable SOA arrays and arrayed waveguide grating routers.

[0058] The working principles of the network card module and CPO-TR module using photoelectric conversion technology are as follows: Figure 2 As shown, to achieve high-speed Ethernet frame switching of the ONet module, four 800Gbps network card modules and four CTO-TR optical modules were developed. The transmitting part of the optical module mainly includes a DSP, TOSA (tunable laser, etc.), LDD (optical modulator), and MUX (wavelength multiplexer); the receiving part mainly includes a DSP, TIA / LA (current converter / current limiting amplifier), PD (photodetector), and DeMux (wavelength demultiplexer). The transmitting and receiving parts each use 16 optical fibers to carry optical waves of different wavelengths. Each wave is modulated with a 224Gbps optical carrier signal using PAM4 / 8 / 16 modulation format as needed. The biggest difference from traditional optical modules is the use of tunable lasers, which are set to output 16 different wavelengths of fundamental waves.

[0059] The working principle of a programmable SOA array module using high-speed optical switching technology is as follows: Figure 3 As shown, the injected current of the SOA enables two states: "on" and "off." The optical signal will obtain high gain and loss respectively. By controlling the operating current and magnitude of the SOA, the optical switching function and high-speed switching rate of the SOA can be achieved. As a core component, the SOA is transparent to bit rate, modulation method, and optical signal; it has a nanosecond-level switching rate; it has no switching loss and can even have gain; it is flexible in networking and easy to integrate with silicon photonics.

[0060] To enable the SOA to function as an optical switch, a switching current pulse must be applied. To achieve a high switching speed, the rise and fall times of the applied switching current pulse must be very short, and it must have a large current drive capability. This requires a high-speed, high-current drive circuit unit within the SOA array. To ensure reliable SOA operation, temperature control is essential. Encapsulating a thermistor and TEC cooler within the SOA module enables temperature control. Therefore, a high-performance and stable temperature control circuit unit is required within the SOA array. The gain of the SOA is not stable, varying with device temperature and also related to the input optical power. This characteristic affects the use of SOA-based optical switching systems. To improve the dynamic range and stabilize the SOA's output optical power, gain control is necessary. This requires a gain control circuit unit within the SOA array. The specific circuit designs for the high-speed, high-current drive circuit unit, temperature control circuit unit, and gain control circuit unit refer to industry-standard design methods and will not be elaborated here.

[0061] The programmable SOA array communicates with the switching controller via a PCIe3 bus. Therefore, a PCIe3 bridge is set up, and the internal bus is interconnected to the main control module. The main control module completes command interaction with the switching controller through the PCIe3 bridge as needed. The main control module is also responsible for SOA driving, temperature control, and gain control. Through the SOA high current drive unit, there are internal control lines connected to each SOA unit. The main control module realizes high-speed opening and closing of SOA units by controlling the current output of the current drive unit.

[0062] like Figure 3 As shown, the SOA array section of the programmable SOA array module mainly consists of 16 1×16 splitters, 16 SOA groups (each group containing 16 SOA), and 16 16×1 optical couplers. Each splitter's 16 outputs are interconnected one-to-one with a group of 16 SOA, and each SOA group's outputs are interconnected with the 16 16×1 optical couplers using a CLOS topology. Finally, the 16 optical outputs of the optical couplers are achieved by controlling the control signals output from the main controller to the SOA, which form the control code.

[0063] By constructing a programmable SOA array module and working in conjunction with the AWGR-T module, parallel wavelength horizontal queue multi-port input and single-port wavelength vertical queue output can be achieved. This technical solution only requires the implementation of IP data layer 2 forwarding and SOA collaborative switching control. Compared with the traditional method of single-wavelength independent information carrying and time-slot control switching, it does not require high-precision clock synchronization or fine-grained time slot scheduling. It is simple, practical and easy to commercialize. To build a high-speed optoelectronic collaborative data exchange system, it is necessary to construct a new working mechanism for the control plane of optoelectronic collaborative switching. This mechanism mainly includes: interactive communication between the switching controller and the switching agent based on PCIe3 x4, communication and control between the switching controller and the SOA array based on the PCIe3 x4 bus, the collaborative working mechanism between the control plane and the data plane, and the update mechanism of the OWS optical channel mapping table.

[0064] 1) Implementation of interactive communication between the switching controller and the switching agent based on PCIe 3×4 like Figure 4 As shown, the switching controller, acting as the master device, is interconnected with the switching agent, which acts as the slave device, via a PCIe 3×4 bus. To achieve cooperative optoelectronic switching, a series of registers are defined as follows: Switch agent-side status register settings:

[0065] Switching agent-side MAC data register settings:

[0066] 2) Communication and control implementation between the switching controller and SOA array based on PCIe 3×4 bus like Figure 5 As shown, the switching controller, acting as the master device, is interconnected with each SOA switch array of the OWS module, which acts as a slave device, via a PCIe 3×4 bus. To achieve cooperative optoelectronic switching, the optical channel switching register is defined as follows:

[0067] 3) The collaborative working mechanism between the control plane and the data plane: like Figure 6 The image illustrates the process of sending an Ethernet data frame. Figure 1 The electrical switching module distinguishes between internal and external data processing. MAC addresses identify internal frames for processing, and internal switching handles forwarding on ports ①②③④. If a frame is identified as an internal frame, it is forwarded via port 5. Through the PCIe bus on port ⑤ and the DMA mechanism of the PCIe interface controller of the switching agent, the Ethernet frame is buffered in the switching agent's data buffer. Upon receiving an Ethernet frame, the switching agent divides it into four equal parts as data information and synchronously sends an RTS request to the switching controller. The switching controller, after determining the availability of the optical channel, replies with a CTS signal to the switching agent. The switching agent then sends the storage addresses of the four data information parts to the RDMA network card module and commands it to send. The network card module, through the already deployed CPO-TR module and optical channel link, directly completes data transmission with the peer network card module using RDMA technology.

[0068] like Figure 7The diagram illustrates the process of receiving an Ethernet data frame. The network interface card (NIC) module at the other end receives four data packets in parallel, buffers them in the buffer of the switching agent, reassembles them into a single data frame according to their original order, and sends them to port 5 of the local electrical switching module. Simultaneously, it notifies the switching controller that the reception is complete via the LFRC register.

[0069] 3) OWS optical channel mapping table update mechanism: As mentioned earlier, the switching controller can open optical channels in response to requests from the switching agent, but it needs to know which optical channel to open—that is, how the switching controller knows which switch code to send to the programmable SOA switch array. We know that electrical switching modules operate according to MAC address learning and aging mechanisms. Here, we define an OWS optical channel mapping table, which is used to map MAC addresses to OWS SOA switch array codes. The switching controller finds the corresponding SOA switch array code by looking up this table.

[0070] OWS optical channel mapping table of the switching controller:

[0071] Ethernet frames on port 5 of each switching module are processed and transmitted transparently by the switching agent. Broadcast frames are also transmitted and transmitted across all optical channels. However, during the transmission of unicast frames, it is necessary to know which optical channel the frame to be switched needs to be transmitted through. This problem is solved through the optical channel mapping table update mechanism. 1. MAC Address Table Update Trigger Mechanism: The switching controller collects and parses the MAC address table M5T(n) of port 5 of each ONET switching agent / switching module n. Regardless of whether it is a MAC table learning mechanism or an aging mechanism, any update will trigger an update of the MAC table set on the switching controller side: {(M5T(n)), ..., (M5T(m))}. 2. Update Mechanism: Compare the {M5T(n)} of each ONET. Those with the same MAC address are considered valid port channel matches. The corresponding interconnecting OWS optical link channel address code is written into an OWS optical channel mapping table. Through the above update mechanism, we can obtain an OWS optical channel mapping table, but this also brings new problems: time conflicts and efficiency issues between the request for use of the optical channel mapping table and the ongoing table update. For example... Figure 8 As shown, we solve this problem using the following method: 1. The mapping sub-table is updated in real time according to the above mechanism, and the estimated time to complete one round of update is no more than 500ns; the mapping master table refreshes and synchronizes the changes in the mapping sub-table at regular intervals (e.g., every 10us).

[0072] 2. Frame switching requests an optical channel. First, look up the OWS optical channel mapping master table. If the corresponding MAC address and optical channel address code are found, the process ends.

[0073] 3. If the switching controller cannot find the address, it is allowed to refresh the mapping sub-table to the mapping master table immediately, search again, and if successful, the process ends.

[0074] 4. If unsuccessful, notify the exchange agent to resend until successful.

[0075] System basic operation process With the aforementioned optoelectronic collaborative high-speed data exchange device and optoelectronic collaborative exchange control device, the optoelectronic integrated data exchange system has a data plane and a control plane, as well as a collaborative working mechanism between the two. Currently, only the basic operation process of a logically self-consistent system remains to be addressed.

[0076] Figure 9 The basic flow of the control plane of the optoelectronic converged data exchange system is explained, which mainly includes the execution flow of SOA array routing of the switch controller to OWS, the execution flow of MAC table reporting by the switch agent, the update flow of the OWS mapping table of the switch controller, the execution flow of unicast frame routing of the switch controller, and the execution flow of broadcast frame routing of the switch controller.

[0077] Figure 10 The basic flow of the data plane in the optoelectronic converged data exchange system is explained. From the source end, it mainly includes unicast frame transmission flow and broadcast frame transmission flow.

[0078] The specific application scenarios of this embodiment are as follows: Scenario 1: The technical components described in this disclosure can be deployed and implemented independently, but in order to further reduce power consumption and improve performance, such as Figure 11As shown, this embodiment prioritizes a distributed optoelectronic co-switching silicon photonic integrated chip deployment. The ONET module is distributed on an optoelectronic co-packaged system-on-a-chip. With the increasing maturity of silicon photonic manufacturing processes and CMOS manufacturing processes, some important silicon photonic devices can be integrated with existing silicon-based chips in 2.5D or 3D form. This makes on-chip optical interconnect a new form of on-chip interconnect. The electrical switching module, switching agent, and network card module belong to the CMOS manufacturing process of electronic devices and can be considered to be integrated on a single chip. The CPO-TR module, as an optical device, can be deployed on an optical integrated circuit. The two are integrated on a carrier board through a silicon interposer using 2.5D or 3D integration technology and used externally as an integrated circuit. The OWS module, which includes a controllable SOA switch matrix and AWGR-T, is also deployed on another system-on-a-chip using silicon photonic integration technology and used as an integrated circuit. The switching controller, due to its flexible and customizable functions, can be deployed on an FPGA device without the need for integrated circuit hardening.

[0079] Scenario 2: such as Figure 12 The diagram illustrates the optoelectronic cross-connect deployment of the communication equipment according to the technical solution disclosed herein. The equipment includes: 16 service boards, one primary and one backup cross-connect board; within each service board: one ONet module is deployed, with each ONet interconnected with one OWS via 16 optical links; each service board provides 4x400G service interfaces externally; within each cross-connect board: one switching controller module and 16 OWS modules are deployed, interconnected in a Dragonfly+ network topology. Each OWS module is allocated one optical channel for each link, carrying 16 different wavelengths of light. The remaining optical channel of each OWS module is looped back for debugging and equipment maintenance. The communication equipment can be a data center switch or a 5G / 6G wide-area network equipment. The overall external service switching capacity of the communication equipment is 25.6Tbps.

[0080] Scenario 3: such as Figure 13 The diagram illustrates the deployment of the disclosed technical solution for optoelectronic cross-connect capacity expansion in communication equipment. As described above, this patent employs a 4×400Gbps electrical switching module, a 4×800Gbps network card module, a 16×224Gbps CPO-TR module, a 16×16 programmable SOA array, and a 16×16 AWGR-T module. However, to expand the switching capacity of communication equipment, a 4×800Gbps electrical switching module, an 8×800Gbps network card module, a 32×224Gbps CPO-TR module, a 32×32 programmable SOA array, and a 32×32 AWGR-T module can also be used; the working principle remains the same. Figure 13As shown on the right, the communication equipment is deployed as follows: 32 service boards, with one primary and one backup cross-connect board (4 boards in total); within each service board: one ONet module is deployed, and one ONET is interconnected with one OWS via 32 optical links. Each service board provides 4 x 800G Ethernet service interfaces externally; within each cross-connect board: one switching controller module and 16 OWS modules are deployed. The 32 OWS modules are directly connected in a Dragonfly+ network topology. Each OWS module is allocated one optical channel for each link, and each channel carries 32 different wavelengths of light. The remaining optical channel of the OWS module is looped back for debugging and equipment maintenance. The overall external switching capacity of the communication equipment can be expanded to 102.4Tbps.

[0081] Scenario 4: such as Figure 14 As shown, to adapt to low-switching capacity scenarios and reduce equipment costs, 4×200Gbps electrical switching modules, 4×400Gbps network card modules, 8×224Gbps CPO-TR modules, 8×8 programmable SOA arrays, and 8×8 AWGR-T modules can also be used, with the same working principle. The equipment deployment consists of 8 service boards, with 1 primary and 1 backup cross-connect board. Within each service board, one ONet module is deployed, and one ONET is interconnected with one OWS via 8 optical links. Each service board provides 4x200Gbps Ethernet service interfaces. Within each cross-connect board, one switching controller module and 8 OWS modules are deployed. The 8 OWS modules are directly connected in a Dragonfly+ topology. Each OWS module is allocated one optical channel for each link, carrying 8 different wavelengths of light. The remaining optical channel of the OWS module is looped back for debugging and equipment maintenance. The overall external service switching capacity of the communication equipment can be reduced to 6.4Tbps.

[0082] The system's hardware architecture can be implemented using a bus architecture. A bus architecture can include any number of interconnect buses and bridges, depending on the specific application and overall design constraints of the hardware. A bus connects various circuits, including one or more processors, memory, and / or hardware modules. It can also connect various other circuits such as peripherals, voltage regulators, power management circuits, external antennas, etc. Buses can be Industry Standard Architecture (ISA) buses, Peripheral Component Interconnect (PCI) buses, or Extended Industry Standard Component (EISA) buses, etc. Buses can be categorized as address buses, data buses, control buses, etc.

[0083] In the description of this specification, the references to terms such as "one embodiment / mode," "some embodiments / modes," "example," "specific example," or "some examples," etc., refer to specific features, structures, or characteristics described in connection with that embodiment / mode or example, which are included in at least one embodiment / mode or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment / mode or example. Moreover, the specific features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments / modes or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments / modes or examples described in this specification, as well as the features of different embodiments / modes or examples.

[0084] Those skilled in the art should understand that the above embodiments are merely for illustrating the present disclosure and are not intended to limit the scope of the disclosure. Those skilled in the art can make other changes or modifications based on the above disclosure, and these changes or modifications still fall within the scope of the present disclosure.

Claims

1. A photoelectric collaborative high-speed data exchange system, characterized in that, It includes an electrical processing unit, an optical switching unit, and an optical switching controller; among which, The electrical processing unit is configured to: receive a data frame, divide the data frame into multiple cell slices, and allocate a buffer address to each cell slice; The optical switching controller is configured to: receive a request to send a message, query wavelength scheduling configuration information based on the destination address information in the request to send the message, and generate wavelength scheduling control instructions; The optical switching unit includes a controllable optical switch array and an arrayed waveguide grating router. The optical switching unit is configured to: receive the wavelength scheduling control command, control the conduction state of the controllable optical switch array according to the wavelength scheduling control command, and establish an optical channel from the source end to the destination end by utilizing the wavelength routing characteristics of the arrayed waveguide grating router. The electrical processing unit also integrates a remote direct memory access unit, which is configured to: after receiving a message to allow transmission, output the cache address and data length of each cell slice, so that the destination electrical processing unit can directly read the cell slice from the source cache based on the cache address and data length and reassemble it into a data frame.

2. The optoelectronic collaborative high-speed data exchange system according to claim 1, characterized in that, The electrical processing unit includes an electrical switching function module and a control agent; the control agent is configured to process the data frame and, when determining that the data frame is a broadcast frame, send a broadcast request message to the optical switching controller.

3. The optoelectronic collaborative high-speed data exchange system according to claim 1, characterized in that, The remote direct memory access unit is configured to output the cache address and data length of each cell slice via a high-speed serial bus.

4. The optoelectronic collaborative high-speed data exchange system according to claim 1, characterized in that, The optical switching unit is connected to the electrical processing unit via a photoelectric conversion interface; the photoelectric conversion interface is configured to: convert electrical signals into multiple optical signals and transmit them to the optical switching unit via a transmitting optical channel, and receive multiple optical signals via a receiving optical channel and convert them into electrical signals; the transmitting optical channel and the receiving optical channel are respectively used for unidirectional optical transmission of data.

5. The optoelectronic collaborative high-speed data exchange system according to claim 4, characterized in that, The controllable optical switch array is implemented using a semiconductor optical amplifier; the semiconductor optical amplifier is configured to control its on or off state by injecting current, and has a nanosecond-level switching rate; the optical switching unit also includes a high-speed high-current drive circuit unit, a temperature control circuit unit, and a gain control circuit unit.

6. The optoelectronic collaborative high-speed data exchange system according to claim 1, characterized in that, The optical switching controller stores two sets of wavelength scheduling configuration information, one of which is in an active state and the other is in an inactive state. The optical switching controller is configured to perform batch updates on the wavelength scheduling configuration information in the inactive state, and switch the wavelength scheduling configuration information to the active state after the update is completed.

7. The optoelectronic collaborative high-speed data exchange system according to claim 2, characterized in that, The optical switching controller is configured to: upon receiving the broadcast request message, return a broadcast permission message to the electrical processing unit; the optical switching unit is configured to: upon receiving the broadcast permission message, perform broadcast pass-through through all optical channels.

8. The optoelectronic collaborative high-speed data exchange system according to claim 1, characterized in that, Multiple optical switching units are interconnected according to the Dragonfly+ topology.

9. The optoelectronic collaborative high-speed data exchange system according to claim 1, characterized in that, The system is deployed in a cross-connection scenario of communication equipment and includes multiple service boards and at least one cross-connect board; the cross-connect board contains the optical switching unit and the optical switching controller.

10. A communication device, characterized in that, Including the optoelectronic collaborative high-speed data exchange system as described in any one of claims 1 to 9.

Images