A method and system for dynamic queue management of multi-priority services in time-sensitive network switching

By configuring independent queue lists and shared cache resources for time-sensitive network switching, combined with strict priority scheduling, the problems of high hardware resource consumption and low multicast forwarding efficiency in time-sensitive networks are solved, realizing low-latency and high-reliability transmission of critical services and efficient data forwarding in multicast scenarios.

CN123123286APending Publication Date: 2026-11-17XIAN MICROELECTRONICS TECH INST
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202611475668.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies in time-sensitive network switching suffer from problems such as high hardware resource consumption, weak port expansion capabilities, low multicast forwarding efficiency, and inability to adapt to multi-priority service scheduling, resulting in a decline in network transmission performance and an inability to meet the high-speed, high-concurrency data transmission requirements in complex scenarios.

Method used

A multi-priority dynamic queue management method is adopted. By independently configuring queue linked lists for data frames of different priorities for each destination port, combined with shared cache resources and strict priority scheduling, isolated storage and scheduling of data frames are achieved. This supports dequeuing of multicast frames in stages and unified updates, thus optimizing the queue management process.

Benefits of technology

It achieves low-latency and high-reliability transmission for critical services, improves data forwarding efficiency in multicast scenarios, saves hardware cache resources, adapts to real-time network application scenarios of different scales, and has good scalability and high cache utilization.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123286A_ABST
    Figure CN123123286A_ABST
Patent Text Reader

Abstract

This invention relates to the field of time-sensitive network communication control technology, and discloses a method and system for dynamic queue management of multi-priority services in time-sensitive network switching. The method initializes queue lists and shared cache resources, constructs a multi-port, multi-priority service queue storage architecture, configures separate queue lists for data frames of different priorities for each destination port, and divides the shared cache into standardized cache blocks with corresponding management structures. It receives data frames, extracts priority and forwarding port information, and maps them to preset priority levels. It matches queue lists according to priority and port, completes data frame enqueueing based on queue status, and updates queue and cache information. It performs dequeueing based on strict priority scheduling, reads data frames to complete port transmission, and releases the cache and updates the idle state after transmission. For multicast frames, it adopts a method of dequeuing in stages and updating the queue status uniformly to achieve dynamic queue management. This invention improves the real-time transmission capability of services.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of time-sensitive network communication control technology, specifically to a method and system for dynamic queue management of multi-priority services in time-sensitive network switching. Background Technology

[0002] In the field of real-time network communication, Time-Sensitive Networking (TSN) and Time-Triggered Ethernet (TTE) are widely used in scenarios with stringent requirements for data transmission performance, such as industrial control, aerospace, and automotive networks, due to their high-precision clock synchronization, low latency, low jitter, and high-reliability transmission characteristics. The real-time performance and reliability of network data transmission are the core keys to ensuring the stable operation of the overall system.

[0003] Currently, existing technologies include design schemes for queue scheduling and bus control of TTE-ET services. These schemes set up independent switching planes for TT time-triggered services and ET event-triggered services, and employ a dedicated queue management mechanism for ET services to differentiate the transmission scheduling logic of the two types of services and adapt to the transmission requirements of different services. However, this technical solution has significant drawbacks in practical applications. On the one hand, the architecture design of an independent switching plane with dedicated queue management consumes a great deal of FPGA hardware resources, which limits the expansion capability of switch ports and makes it difficult to meet the deployment requirements of multi-port, large-scale real-time networks. On the other hand, the existing solution has an imperfect forwarding mechanism for multicast frames, which significantly reduces the data forwarding efficiency of the switch in multicast data transmission scenarios, leading to a decline in the overall network transmission performance and making it unable to adapt to the high-speed, high-concurrency data transmission requirements in complex scenarios.

[0004] The existing patented invention CN102308537A proposes a queue management scheme based on queue frame counting and token count. It achieves queue control by counting queue frames and matching token counts, which can regulate the order of queue data transmission to some extent. However, this patented solution has a relatively simple design, only implementing basic queue traffic control. It does not design differentiated processing mechanisms for packets with different priorities in the network, making it unsuitable for mixed transmission scheduling scenarios with multiple priorities and types of services in TTE and TSN networks. It struggles to solve problems such as unstable latency caused by mixed transmission of high and low priority packets and the inability to guarantee the real-time performance of high-priority services, resulting in poor applicability and compatibility.

[0005] In summary, current real-time network queue management and scheduling technologies generally suffer from high hardware resource consumption, weak port expansion capabilities, low multicast forwarding efficiency, and inability to adapt to multi-priority service scheduling. They are unable to simultaneously meet the real-time performance, reliability, scalability, and multi-service adaptability of real-time networks. Therefore, a new queue management solution is urgently needed to address the numerous problems existing in the current technologies. Summary of the Invention

[0006] In order to overcome the shortcomings of the existing technology, the purpose of this invention is to provide a method and system for dynamic queue management of multi-priority services in time-sensitive network switching, so as to solve the technical problem of queuing and blocking of critical traffic within the switching node during real-time network communication.

[0007] This invention is achieved through the following technical solution: In a first aspect, the present invention provides a method for dynamic queue management of multi-priority services in time-sensitive network switching, comprising: Initialize the queue linked list and shared cache resources to build a queue storage architecture that adapts to multi-port and multi-priority services. The queue linked list is independently configured for different priority data frames of each destination port. The shared cache resources are divided into standardized cache blocks and matched with corresponding cache management structures. Receive the data frame to be transmitted, extract the priority information and forwarding port information of the data frame, and map the priority information to the system's preset priority level; Based on the mapped priority level and forwarding port information, match the corresponding queue list, perform data frame enqueue operation in combination with the queue idle state, and update the queue list and shared buffer status information. Based on a strict priority scheduling principle, dequeue scheduling is performed on each queue linked list. Data frames are read from the queue linked list according to the scheduling result and port transmission is completed. After the data frame transmission is completed, the corresponding shared buffer resources are released and the status of the idle buffer is updated. For multicast frames, a processing logic of dequeuing in stages and finally updating the queue status uniformly is adopted to complete the dynamic queue full-process management.

[0008] Preferably, the specific process of initializing the queue linked list and shared cache resources includes: Initialize the head and tail pointers of each priority queue linked list corresponding to each destination port to complete the basic architecture of the queue linked list; The shared cache data cache is divided into 64B fixed cache blocks, and each fixed cache block can store a minimum data frame; Configure cache block descriptors that correspond one-to-one with each cache block, and a cache of free block addresses used to record the usage and release status of free cache blocks, thus completing the initialization of shared cache resources.

[0009] Furthermore, the information recorded in the cache block descriptor includes the start and end flags of the data frame corresponding to the data fragment stored in the cache block, the output port, the effective data length, and the address information of the next cache block.

[0010] Preferably, the priority information extraction method for the data frame includes: Extract the 3-bit PCP field of VLAN priority from the 802.1Q tag of the data frame as priority information; Generate priority information according to the data frame type; The eight VLAN priorities are mapped to the system's preset high, medium, and low priority levels.

[0011] Furthermore, the priority mapping rule is as follows: VLAN priority 7 corresponds to high system priority, VLAN priorities 0, 1, 2, and 3 correspond to low system priority, and the remaining VLAN priorities correspond to the system priority.

[0012] Preferably, the specific process of matching the corresponding queue linked list based on the mapped priority level and forwarding port information is as follows: The current queue status is checked. If the queue is not full, the data frame is directly linked to the tail of the queue. If the queue is full, the lowest priority data frame in the current queue is discarded to free up storage space for higher priority data frames. A flow control signal is sent to the data frame sender to slow down the data transmission rate. When a data frame is allowed to be enqueued, the corresponding number of free buffer blocks are matched according to the data frame length and the data is stored. The address of the next free buffer block is matched according to the buffer block descriptor to complete the enqueueing of the entire frame.

[0013] Preferably, the status information of the update queue linked list and shared cache includes the update queue frame count, queue head and tail pointers, and the occupancy and release information of the free cache block address cache.

[0014] Preferably, the strict priority scheduling principle prioritizes scheduling data frames from high-priority queues. The scheduling output information includes the output port, priority, and multicast flag information. Based on the scheduling information index, the corresponding dequeue queue is read, the queue head pointer is read, and the buffer block data is read sequentially until all data reading and port transmission of a single data frame are completed.

[0015] Preferably, the specific process of the processing logic for multicast frames, which involves dequeuing in stages and finally updating the queue state uniformly, includes: During multicast frame scheduling, if some output ports are not idle, only the currently idle output ports are selected as the scheduling result to complete the dequeue transmission of some port data. After all multicast frames have been transmitted across all ports, release the storage space of all corresponding buffer blocks and update the free block address cache and the head pointer of the multicast queue.

[0016] Secondly, the present invention also provides a multi-priority service dynamic queue management system for time-sensitive network switching, comprising: The resource initialization module is used to initialize the queue linked list and shared cache resources, and build a queue storage architecture that adapts to multi-port and multi-priority services. The queue linked list is independently configured for different priority data frames of each destination port, and the shared cache resources are divided into standardized cache blocks and matched with corresponding cache management structures. The priority parsing module is used to receive data frames to be transmitted, extract the priority information and forwarding port information of the data frames, and map the priority information to the system's preset priority level. The queue enqueue module is used to match the corresponding queue list based on the mapped priority level and forwarding port information, perform data frame enqueue operation in combination with the queue idle state, and update the status information of the queue list and shared buffer. The queue dequeue module is used to perform dequeue scheduling on each queue linked list based on a strict priority scheduling principle, and read data frames from the queue linked list according to the scheduling result and complete port transmission. The resource reclamation module is used to release the corresponding shared buffer resources and update the status of idle buffers after the data frame transmission is completed. For multicast frames, it adopts a processing logic of dequeuing in stages and finally updating the queue status uniformly to complete the dynamic queue full-process management.

[0017] Compared with the prior art, the present invention has the following beneficial technical effects: This invention provides a dynamic queue management method for multi-priority services in time-sensitive network switching. By independently configuring queue lists for data frames of different priorities at each destination port, it implements isolated storage and independent scheduling of data with different priorities. This avoids queuing congestion caused by low-priority traffic preempting queuing resources for critical service traffic in real-time networks, ensuring priority scheduling and forwarding of critical service packets, improving the transmission efficiency of critical services, and meeting the real-time transmission requirements of time-sensitive networks. This queue list architecture can be expanded by adding corresponding queues as the number of switch ports and internal priorities increase, exhibiting good scalability and adaptability to real-time network application scenarios of different scales, demonstrating strong versatility. For multicast frames, it adopts a processing method of dequeuing in stages and updating the queue status uniformly after all ports have completed transmission. It does not require waiting for all forwarding ports of the multicast frame to be in an idle state before starting forwarding, supporting multiple scheduling and forwarding of multicast frames, and improving data forwarding efficiency in multicast scenarios. At the same time, it adopts a multi-path data aggregation combined with a shared caching mechanism, dividing the shared cache into standardized cache blocks for unified management, efficiently reusing cache storage space, saving hardware cache resources, and maximizing the utilization of link bandwidth. Attached Figure Description

[0018] Figure 1 This is a flowchart of a multi-priority service dynamic queue management method for time-sensitive network switching in an embodiment of the present invention; Figure 2This is a cache block descriptor information diagram in an embodiment of the present invention; Figure 3 This is a free block address cache diagram in an embodiment of the present invention; Figure 4 This is a schematic diagram of the 802.1Q tag priority field in an embodiment of the present invention; Figure 5 This is a schematic diagram of the frame type field in an embodiment of the present invention; Figure 6 This is a block diagram of the queue management controller in an embodiment of the present invention; In the diagram: 1. Port polling module; 2. Enqueue control module; 3. Shared cache management unit; 4. Data cache unit; 5. Cache block descriptor cache unit; 6. Free block address cache unit; 7. Dequeue control unit; 8. Multi-priority queue linked list management unit. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0020] The purpose of this invention is to provide a method and system for dynamic queue management of multi-priority services in time-sensitive network switching. It establishes a multi-port, multi-priority queue storage and cache management architecture with low hardware resource consumption, high scalability, and good isolation. This architecture avoids the technical defects of traditional queue management at the underlying hardware architecture level, providing stable, reliable, and scalable hardware support for the entire process of subsequent data frame orderly enqueueing, fine-grained scheduling, efficient dequeueing, and closed-loop resource recycling. This ensures the real-time performance and reliability of time-sensitive network service transmission, thereby solving the technical problem of queuing and congestion that may occur in critical traffic within switching nodes during real-time network communication.

[0021] The present invention will now be described in further detail with reference to the accompanying drawings: Example 1 See Figure 1 In one embodiment of the present invention, a dynamic queue management method for multi-priority services in time-sensitive network switching is provided. The overall hardware logic is implemented based on the queue management controller designed in this invention, and the structure of the queue management controller is as follows: Figure 6As shown in the attached diagram, the corresponding labels for each hardware unit are as follows: Port Polling Module 1, Enqueue Control Module 2, Shared Cache Management Unit 3, Data Cache Unit 4, Cache Block Descriptor Cache Unit 5, Free Block Address Cache Unit 6, Dequeue Control Unit 7, and Multi-Priority Queue Linked List Management Unit 8. The entire controller relies on a hardware pipeline architecture comprised of the Multi-Priority Queue Linked List Management Unit 8, Shared Cache Management Unit 3, Port Polling Module 1, Enqueue Control Module 2, and Dequeue Control Unit 7. Combined with core hardware logic such as 512-bit fragmented read / write, dynamic queue pointer update, priority mapping, congestion flow control, strict priority scheduling, and multicast frame XOR judgment for dequeueing, it fully implements dynamic queue management for multi-priority services, adapting to the high-speed parallel processing characteristics of FPGAs. The detailed steps and corresponding hardware execution flow are as follows: Step 1: Initialize the queue linked list and shared cache resources to build a queue storage architecture that adapts to multi-port and multi-priority services. The queue linked list is independently configured for different priority data frames of each destination port. The shared cache resources are divided into standardized cache blocks and matched with corresponding cache management structures. Specifically, regarding queue list initialization, this invention employs a unique multi-priority independent queue list management mechanism. It abandons the traditional shared queue and mixed storage / scheduling design, configuring a dedicated queue for each priority data frame on each destination port of the switch. All queues are uniformly maintained and managed through a standardized linked list structure. During power-on initialization, the system automatically initializes the queue list parameters for all destination ports and priorities across the entire network. It focuses on resetting and initializing the head and tail pointers of each queue list, establishing a well-organized and independent multi-queue storage infrastructure. Through the isolated configuration of dedicated queue lists, it completely achieves physical storage isolation and independent scheduling of data frames with different priorities, fundamentally eliminating problems such as queuing contention, critical service blockage, and scheduling disorder caused by mixed storage of high and low priority data. Meanwhile, this queue linked list architecture has strong horizontal and vertical scalability, and high engineering adaptability and versatility. When the number of physical ports of the switch increases, the network scale expands, or the number of internal priorities is added or expanded according to business needs, there is no need to modify the overall hardware architecture, scheduling logic, or cache structure. Only the corresponding number of queue linked lists need to be added to the corresponding modules to complete the rapid adaptation. It can fully meet the deployment needs of real-time networks of different sizes, such as small, medium and large, and solve the technical pain points of traditional FPGA queue architecture, such as limited port expansion, low resource reuse rate and rigid architecture.

[0022] Regarding queue list initialization, this invention employs a multi-priority queue list management mechanism. A dedicated queue is independently configured for each priority data frame on each destination port of the switch, and unified maintenance is achieved through a linked list structure. During the initialization phase, the head and tail pointers of all queue lists are reset and initialized, establishing an isolated storage infrastructure for data frames of different priorities. This enables physical storage isolation and independent scheduling of high and low priority data. This architecture possesses strong scalability. When the number of physical ports on the switch increases or the number of custom priorities within the system expands, no modification to the overall architecture is required; only the corresponding number of queue lists need to be added to adapt, meeting the deployment needs of real-time networks of different scales.

[0023] Regarding shared cache resource initialization, to address the technical problems of traditional time-sensitive network switches such as exclusive use of dedicated queue cache resources, severe cache fragmentation, excessive on-chip FPGA resource consumption, and low bandwidth utilization, this invention adopts a global shared cache unified management strategy to replace the traditional static dedicated cache allocation mode, significantly reducing FPGA hardware resource overhead and improving cache resource utilization. The overall architecture of the shared cache designed in this invention strictly corresponds to the system functional requirements, comprising three main parts: data cache, cache block descriptor cache, and free block address cache. These three parts work collaboratively, each performing its own function, to jointly complete the entire cache management process, including data frame storage, cache status recording, and free resource scheduling. To effectively utilize the limited on-chip cache resources of the FPGA and avoid cache space waste and storage fragmentation, this invention performs equal and refined partitioning of the overall data cache, evenly dividing the entire data cache into standardized 64B fixed-size cache blocks. This size has been optimized through engineering adaptation, and the storage space of a single cache block can precisely accommodate the minimum Ethernet data frame, maximizing adaptation to Ethernet frame storage characteristics and avoiding redundant storage space waste.

[0024] Meanwhile, to achieve precise location, independent management, and chained scheduling of each cache block, this invention matches each physical cache block with a cache block descriptor, constructing a fine-grained management model of "one block, one descriptor." The cache block descriptor information is as follows: Figure 2 As shown. This cache block descriptor serves as the core index information for cache scheduling, accurately recording the start and end flags of the data frame corresponding to the data fragment stored in the current cache block, the data frame output port number, the effective data length within the cache block, and the physical address information of the next free cache block. Through descriptor-based chained indexing, it achieves chained storage and ordered reading of data frames of arbitrary length, ensuring that both long and short frames can be stored appropriately. In addition, the system is configured with an independent free block address cache, such as... Figure 3As shown, this system is specifically designed to record the idle, occupied, and release / reclaimed status of all cache blocks across the entire network in real time, dynamically maintaining a list of idle cache block addresses. This provides accurate address indexes for idle cache allocation during the subsequent data frame enqueue phase and cache resource reclamation and reuse during the dequeue phase, completely resolving the problems of chaotic idle resource statistics, inefficient cache block reuse, and severe resource fragmentation in traditional cache management. Through the above complete initialization configuration, the entire shared cache management structure is finally built, providing efficient, stable, and reusable cache resource support for subsequent dynamic queue management.

[0025] Step 2: Receive the data frame to be transmitted, extract the priority information and forwarding port information of the data frame, and map the priority information to the system's preset priority level; Specifically, this step is the core of the pre-processing parsing and priority standardization for data frame enqueueing, playing a crucial role in multi-channel data aggregation, frame key information parsing, and unified mapping of heterogeneous priorities. In complex time-sensitive network scenarios, the network simultaneously carries multiple types of data, including real-time critical control services, routine transmission services, and general background services. The priority identifiers for different services come from different sources and have inconsistent level standards. There are both VLAN priorities defined by standard protocols and service-defined exclusive priorities. If scheduling is directly based on native priorities, it is highly likely to lead to problems such as chaotic scheduling standards, incorrect service priorities, and inadequate protection of critical services. This step, through standardized parsing and unified mapping mechanisms, completes the aggregation, filtering, and priority normalization of multi-channel heterogeneous data, achieving unified hierarchical management of priority data from different protocols, sources, and specifications. This provides a unified and standardized basis for subsequent differentiated queue matching, accurate enqueueing, and priority scheduling, ensuring that the network-wide service scheduling logic is unified, orderly, and controllable.

[0026] In practical networking applications, switches are configured with multiple gigabit Ethernet ports. These ports can receive multiple data frames of different types and priorities in parallel, leading to scenarios where multiple ports receive frames simultaneously and data floods in concurrently. If multiple data streams are written to the queue at the same time, it can easily cause data conflicts, frame corruption, and write anomalies. To address this, this invention relies on the port polling module 1 in the queue management controller to perform ordered aggregation and filtering of multiple concurrent data streams. Through a hardware polling scheduling mechanism, one valid data stream is selected in a time-sharing manner to enter the subsequent queuing process, effectively avoiding the problem of concurrent write conflicts from multiple ports, ensuring the orderliness and integrity of data frame queuing, and achieving aggregation and convergence of multiple data streams. For each data frame that has completed aggregation and filtering and is ready for processing, the system accurately parses the frame header information. The multi-priority queue linked list management unit 8 parses the frame information of the enqueued data to accurately obtain key information such as the output port number and frame priority, providing core parameter support for subsequent queue matching, priority mapping, and hierarchical scheduling. The priority information of this invention supports dual-mode acquisition, flexibly adapting to standardized protocol scenarios and customized business scenarios, exhibiting strong compatibility and versatility: the first mode is the standard protocol priority extraction mode, adapted to general Ethernet real-time transmission scenarios, accurately parsing the 3-bit PCP field inside the 802.1Q tag of the data frame, such as... Figure 4 As shown, the first mode reads the VLAN priority defined by the protocol standard; the second mode is a custom priority configuration mode, which allows users to manually define exclusive priorities based on the data frame types of different services according to the service needs of special scenarios such as industrial control, aerospace, and vehicle real-time networks, thus meeting the differentiated scheduling needs of non-standard customized services. Figure 5 As shown.

[0027] To eliminate the problems of inconsistent native priority standards, wide range of levels, and high scheduling complexity, and to simplify hardware scheduling logic, reduce FPGA resource consumption, and improve scheduling accuracy, this invention unifies and normalizes the eight fine-grained VLAN priority levels defined by the IEEE 802.1Q protocol into three preset priority queues: high, medium, and low. This achieves streamlined integration and standardized hierarchical classification of multiple priority levels, adapting to the scheduling requirements of prioritizing real-time network core services. This invention sets fixed, strict, and solidifiable priority mapping rules, eliminating dynamic random mapping bias and ensuring accurate and consistent priority determination for each scheduling operation: VLAN priority 7 corresponds to the highest system priority, specifically used to carry real-time critical business data that is latency-sensitive and has extremely high reliability requirements, and is the core business with priority protection across the entire network; VLAN priorities 0, 1, 2, and 3 are uniformly mapped to low system priorities, used to carry ordinary background services with lower latency and reliability requirements, redundant monitoring data, and other non-critical services; the remaining VLAN priorities 4, 5, and 6 are uniformly mapped to high system priorities, used to carry regular important business data, balancing transmission efficiency and resource consumption. Through the standardized mapping process in this step, unified hierarchical management and control of multi-source heterogeneous and multi-specification priority messages are fully realized. This allows business data of different types and sources to correspond to a unique system priority level, providing a precise and unified hierarchical basis for subsequent differentiated queue matching, congestion control, and priority scheduling, and ensuring the standardization and accuracy of scheduling logic from the source.

[0028] Step 3: Based on the mapped priority level and forwarding port information, match the corresponding queue list, perform data frame enqueue operation in combination with the queue idle state, and update the queue list and shared buffer status information; Specifically, this invention achieves intelligent queue management capabilities by implementing a priority-differentiated queuing strategy, a dynamic cache preemption mechanism, and a congestion flow control linkage mechanism. This allows high-priority services to occupy cache resources first, while low-priority services can be dynamically yielded. Under the premise of limited FPGA cache resources, this invention maximizes the transmission quality of critical services and improves overall cache utilization and port throughput.

[0029] Based on the unique destination forwarding port obtained from step two and the normalized system priority level, the system accurately indexes the dedicated queue linked list corresponding to the switch port and priority, achieving one frame per queue and precise matching, thus preventing the problem of mixed storage and switching of packets from different ports and priorities. After matching, the system queries the storage occupancy status and remaining cache space of the current target queue linked list in real time, and executes differentiated enqueueing and congestion control strategies based on the queue's idle status, possessing extremely strong adaptive congestion handling capabilities: if the current target queue's storage space is not full and the remaining cache resources are sufficient, the system directly links the complete data frame to the tail position of the queue linked list, completing the normal enqueueing operation and ensuring fast enqueueing and low-latency queuing of data frames; if the current queue is full and the storage space is completely exhausted, in order to prioritize the transmission performance of real-time critical services and avoid high-priority data being discarded and blocked due to cache overflow, the system actively triggers a dynamic cache preemption mechanism, discarding the lowest-priority data frame currently stored in the queue, and actively freeing up cache storage space for newly accessed high-priority critical data frames. At the same time, the system actively outputs a dedicated flow control signal to the upstream data frame sender to notify the upstream device to reduce the data transmission rate, thereby alleviating the congestion pressure on the switching node from the source and avoiding packet accumulation, increased latency, and batch packet loss caused by continuous congestion, thus achieving a transmission effect of controllable congestion and priority for critical services.

[0030] After a data frame is admitted and enqueued, the system dynamically matches the corresponding number of free buffer blocks based on the actual frame length. Relying on the queue management controller, it completes hardware pipelined fragmented storage. The hardware segments and stores data frames in fixed 512-bit units, adapting to the hardware characteristics of FPGA parallel storage and high-speed read / write, significantly improving data writing efficiency and bus bandwidth utilization. Compared to the one-time storage mode for the entire frame, it effectively adapts to the storage needs of Ethernet frames of different lengths, avoiding the problems of long frame storage stuttering and short frame storage resource waste. The specific hardware execution process is as follows: The multi-priority queue linked list management unit 8 queries the internal queue linked list information based on the mapped queue number to obtain the tail pointer corresponding to the queue. Starting from the tail pointer position, the aggregated frame data is sequentially stored into the data buffer unit 4 in 512-bit units. After each 512-bit data storage segment is completed, the controller reads the physical address of the next free buffer block from the free block address buffer unit 6, and simultaneously writes the complete descriptor information corresponding to the current buffer block into the buffer block descriptor buffer unit 5 in real time, completing the chain binding and information storage of a single data segment. The system continuously executes fragmented write, address update, and descriptor registration operations until all data in the entire frame is written to the shared cache, completing the single-frame enqueue storage operation. This chained cache storage mode based on 512-bit fragmentation can achieve adaptive storage of arbitrary frame lengths, greatly improving the adaptability and resource utilization of the shared cache, and ensuring chained management, ordered storage, and traceable indexing of cache resources.

[0031] After a single frame is completely enqueued and all data fragments are stored, the enqueue control module 2 sends an enqueue completion signal to the multi-priority queue linked list management unit 8, triggering the queue status refresh mechanism. The system immediately and synchronously updates all status information of the global queue linked list and shared cache, ensuring that all system scheduling parameters and cache status are synchronized in real time with accuracy and without deviation. Specific updates include several core parameters: real-time updates of the effective frame count of the target queue linked list, counting the number of data frames currently piled up in the queue; refreshing the head and tail pointer positions of the queue linked list, accurately marking the start and end positions of queue data storage; synchronously refreshing the occupied and released status of the free block address cache unit 6, real-time updating the number and address list of free cache blocks, and accurately recording the occupied, free, and pending release status of all cache blocks, ensuring the accuracy of cache resource scheduling. Through comprehensive status updates, the queue linked list index information, cache resource status, and actual storage situation are completely synchronized, providing accurate data for subsequent dequeue scheduling, resource reclamation, and priority determination. At the same time, the multi-priority queue linked list management unit 8 collects the frame count and accumulation status of each queue in real time, and generates scheduling requests with corresponding weights according to the priority level of different queues. The scheduling requests of high-priority queues have higher priority and response priority, providing the request basis and queue status support for the dequeue control unit 7 to perform strict priority dequeue scheduling.

[0032] Step 4: Based on the strict priority scheduling principle, perform dequeue scheduling on each queue linked list, read the data frame in the queue linked list according to the scheduling result and complete the port transmission; Specifically, this step employs a strictly hardware-fixed priority scheduling principle throughout. The scheduling response priority strictly follows the system's preset high, medium, and low priority levels. The scheduling logic is irreversible, with no unfair preemption or weighted distribution. The system prioritizes responding to scheduling requests from the high-priority queue list. As long as there are data frames to be transmitted in the high-priority queue, the system always prioritizes scheduling high-priority services for dequeue transmission. Only when the high-priority queue has no data to be transmitted and is completely idle will the system respond to scheduling requests from the medium-priority queue in sequence. Similarly, only when both the high and medium priority queues are idle will low-priority queues be scheduled for data transmission. This scheduling logic, from a hardware mechanism perspective, absolutely guarantees the priority dequeueing and forwarding of real-time critical data, completely eliminating the problem of critical services being blocked by low-priority traffic. After completing the scheduling determination, the system outputs standardized scheduling result information, specifically including three core types of information: target output port number, data frame priority level, and frame unicast / multicast flags. This provides a basis for accurate dequeueing and differentiated multicast processing.

[0033] The system accurately indexes the corresponding target queue linked list based on the scheduling result, reads the queue head pointer updated in real time, and, relying on the chained address information and data status information recorded in the cache block descriptor, the dequeue control unit 7 performs the hardware dequeue read operation. For the dequeue process of unicast data frames, the hardware maps the scheduling result to the corresponding queue number, reads the cache block descriptor information stored in the cache block descriptor cache unit 5 from the starting position of the queue head pointer, and reads data fragments sequentially from the data cache unit 4 with a fixed width of 512 bits according to the address, length, and chained index information recorded in the descriptor, outputting the data segment by segment and splicing the complete frame data. The system continues to read cached data until it recognizes that the last cache block of the current data frame has been read, determining that the single frame dequeue is complete, and forwards the spliced ​​valid data frame at high speed to the corresponding destination output port through the switching unit, completing the single data transmission process. After each 512-bit segment of unicast data frame is read, the current cache block resource is immediately released, and the valid address of the released cache block is rewritten to the cache block descriptor cache unit 5 to complete the status write-back. After the entire data frame is dequeued, the head pointer of the corresponding queue and the frame count information in the multi-priority queue linked list management unit 8 are updated synchronously to complete the complete dequeueing and resource reclamation loop of unicast service.

[0034] Step 5: After the data frame transmission is completed, release the corresponding shared buffer resources and update the idle buffer status. For multicast frames, adopt the processing logic of dequeuing in stages and finally updating the queue status uniformly to complete the dynamic queue full-process management.

[0035] Specifically, this invention designs differentiated resource reclamation and queue update logic for unicast and multicast services. Through the innovative mechanism of multicast dequeueing in stages and unified update at the last stage, it completely solves the technical defects of low efficiency in traditional multicast forwarding, and realizes efficient transmission of services and closed-loop management of resources across the entire network.

[0036] For ordinary unicast data frames, the transmission scenario logic is simple, the port is unique, and multiple scheduling is unnecessary. Therefore, after the data frame is completely transmitted and successfully output to the destination port, the system immediately initiates a full resource reclamation process. This process immediately releases all buffer block resources occupied by the current unicast data frame throughout its transmission, rewrites all released free buffer block addresses into the free block address cache, and updates the quantity and address status of free buffers across the entire network in real time. This ensures that buffer resources are reused immediately and without idle waste. Simultaneously, the system synchronously clears the frame count data corresponding to the current queue linked list, refreshes the queue head pointer position in real time, and restores the queue state to the latest free state. This completes the closed-loop processing of the entire unicast service process, ensuring efficient unicast service processing, timely resource reclamation, and accurate queue status.

[0037] For multicast data frames, due to their one-to-many transmission characteristics and multiple output ports, and the fact that the busy and idle states of these output ports are often inconsistent in actual network deployments, making it impossible to guarantee that all ports are idle simultaneously, this invention innovatively adopts a special scheduling management logic of dequeuing in stages and updating on the last update. Relying on a hardware XOR judgment mechanism, it accurately identifies the dequeue count status of multicast frames, adapting to complex scenarios of asynchronous idle multicast ports and significantly improving multicast forwarding efficiency. The specific hardware judgment and execution process is as follows: During the multicast frame dequeue scheduling process, the dequeue control unit 7 performs a hardware XOR operation based on the scheduling result and the preset output port number of the corresponding multicast queue in the multi-priority queue linked list management unit 8. The result of the operation accurately determines whether the dequeue scheduling of this multicast frame is the last dequeue operation for that frame. If the XOR result indicates that this is not the last dequeue operation, the system will only output the data to the currently idle multicast ports to complete the partial forwarding transmission. The system will not update the head pointer and frame count information of the corresponding queue in the multi-priority queue linked list management unit 8, and will retain the complete frame buffer data and the original state of the queue. It will continue to schedule forwarding after the remaining ports become idle. Only when the XOR result indicates that this is the last dequeue operation of the multicast frame, that is, after all destination ports have completed data transmission, will the system release all the buffer block storage space occupied by the multicast frame, update the resource status of the idle block address buffer unit 6 in batches, and finally refresh the head pointer and frame count of the multicast queue linked list. This completes the closed-loop management of the entire multicast service process, ensuring the integrity of multicast data transmission, no packet loss, and no errors, while maximizing the multicast forwarding efficiency and port bandwidth utilization.

[0038] In summary, this invention is based on FPGA hardware architecture design. All logical functions are programmed and module-encapsulated using the Verilog HDL hardware description language. After development, full-function simulation verification is first performed using the Modelsim simulation tool, covering all functional scenarios such as initialization, enqueueing, scheduling, dequeueing, buffer reclamation, multicast processing, and congestion control, verifying the integrity, accuracy, and stability of the logical functions. After successful simulation verification, the project files are downloaded to Xilinx series FPGA hardware devices to complete board-level prototype construction and hardware testing. Simultaneously, a four-port gigabit Ethernet real-time network test environment is built for long-term network function testing and performance stress testing. The test results show that the multi-priority dynamic queue management method designed in this invention operates stably and reliably. Under the four-port gigabit Ethernet aggregation working state of the switch, it can stably receive, process, and forward data frames of different priorities. The data forwarding order strictly follows the preset priority rules, with high-priority critical services always forwarded first, without queuing congestion or excessive latency jitter. There are no abnormalities such as packet loss, packet errors, or out-of-order delivery during the entire network communication process, and the integrity and accuracy of data communication are fully guaranteed. Meanwhile, the system's internal data processing bandwidth fully meets the bandwidth limit requirements of the four Gigabit Ethernet ports, significantly improving bandwidth utilization. This invention, through a multi-priority queue isolation scheduling architecture, completely solves the problem of critical service congestion, ensuring hard real-time transmission performance in time-sensitive networks; through an scalable queue linked list design, it adapts to different network scale scenarios, exhibiting excellent versatility and scalability; through a multicast dequeue mechanism, it significantly improves multicast forwarding efficiency; and through a shared cache aggregation management mode, it effectively saves on-chip cache resources on the FPGA, maximizing link bandwidth utilization. Its overall performance and engineering practicality are significantly superior to traditional queue management solutions, possessing strong engineering feasibility and market application value.

[0039] Example 2 This embodiment also provides a multi-priority service dynamic queue management system for time-sensitive network switching, including: The resource initialization module is used to initialize the queue linked list and shared cache resources, and build a queue storage architecture that adapts to multi-port and multi-priority services. The queue linked list is independently configured for different priority data frames of each destination port, and the shared cache resources are divided into standardized cache blocks and matched with corresponding cache management structures. The priority parsing module is used to receive data frames to be transmitted, extract the priority information and forwarding port information of the data frames, and map the priority information to the system's preset priority level. The queue enqueue module is used to match the corresponding queue list based on the mapped priority level and forwarding port information, perform data frame enqueue operation in combination with the queue idle state, and update the status information of the queue list and shared buffer. The queue dequeue module is used to perform dequeue scheduling on each queue linked list based on a strict priority scheduling principle, and read data frames from the queue linked list according to the scheduling result and complete port transmission. The resource reclamation module is used to release the corresponding shared buffer resources and update the status of idle buffers after the data frame transmission is completed. For multicast frames, it adopts a processing logic of dequeuing in stages and finally updating the queue status uniformly to complete the dynamic queue full-process management.

[0040] In summary, this invention provides a dynamic queue management method and system for multi-priority services in time-sensitive network switching. By optimizing the queue management mechanism, it achieves isolation and independent scheduling of data with different priorities, improving switching and forwarding efficiency while ensuring low latency and high reliability transmission of critical services. This queue management architecture is easily expandable and adaptable to real-time network application scenarios of different scales. This invention improves the transmission efficiency of critical services: by isolating and independently scheduling data with different priorities, it avoids the blocking of critical data by low-priority traffic during real-time network communication, ensuring that critical services can be transmitted quickly and meet real-time requirements. This invention provides a scalable queue management architecture that adapts to real-time network application scenarios of different scales, possessing versatility and strong scalability; this invention supports multiple scheduling and forwarding of multicast frames without requiring all forwarding ports corresponding to multicast frames to be idle before forwarding, further improving data transmission efficiency. This invention employs port aggregation and shared cache management methods to save cache resources and maximize bandwidth utilization.

[0041] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for dynamic queue management of multi-priority services in time-sensitive network switching, characterized in that, include: Initialize the queue linked list and shared cache resources to build a queue storage architecture that adapts to multi-port and multi-priority services. The queue linked list is independently configured for different priority data frames of each destination port. The shared cache resources are divided into standardized cache blocks and matched with corresponding cache management structures. Receive the data frame to be transmitted, extract the priority information and forwarding port information of the data frame, and map the priority information to the system's preset priority level; Based on the mapped priority level and forwarding port information, match the corresponding queue list, perform data frame enqueue operation in combination with the queue idle state, and update the queue list and shared buffer status information. Based on a strict priority scheduling principle, dequeue scheduling is performed on each queue linked list. Data frames are read from the queue linked list according to the scheduling result and port transmission is completed. After the data frame transmission is completed, the corresponding shared buffer resources are released and the status of the idle buffer is updated. For multicast frames, a processing logic of dequeuing in stages and finally updating the queue status uniformly is adopted to complete the dynamic queue full-process management.

2. The method for dynamic queue management of multi-priority services in time-sensitive network switching according to claim 1, characterized in that, The specific process of initializing the queue linked list and shared cache resources includes: Initialize the head and tail pointers of each priority queue linked list corresponding to each destination port to complete the basic architecture of the queue linked list; The shared cache data cache is divided into 64B fixed cache blocks, and each fixed cache block can store a minimum data frame; Configure cache block descriptors that correspond one-to-one with each cache block, and a cache of free block addresses used to record the usage and release status of free cache blocks, thus completing the initialization of shared cache resources.

3. The method for dynamic queue management of multi-priority services in time-sensitive network switching according to claim 2, characterized in that, The information recorded in the cache block descriptor includes the start and end flags of the data frame corresponding to the data fragment stored in the cache block, the output port, the effective data length, and the address information of the next cache block.

4. The method for dynamic queue management of multi-priority services in time-sensitive network switching according to claim 1, characterized in that, The priority information extraction method for the data frame includes: Extract the 3-bit PCP field of VLAN priority from the 802.1Q tag of the data frame as priority information; Generate priority information according to the data frame type; The eight VLAN priorities are mapped to the system's preset high, medium, and low priority levels.

5. A method for dynamic queue management of multi-priority services in time-sensitive network switching according to claim 4, characterized in that, The priority mapping rule is as follows: VLAN priority 7 corresponds to high system priority, VLAN priorities 0, 1, 2, and 3 correspond to low system priority, and the remaining VLAN priorities correspond to the system priority.

6. The method for dynamic queue management of multi-priority services in time-sensitive network switching according to claim 1, characterized in that, Based on the mapped priority level and forwarding port information, the specific process of matching the corresponding queue linked list is as follows: Check the current queue status of the linked list. If the queue is not full, link the data frame directly to the tail of the queue. If the queue is full, discard the lowest priority data frame in the current queue to make room for high priority data frames, and send a flow control signal to the data frame sender to slow down the data transmission rate. When a data frame is allowed to be enqueued, the corresponding number of free buffer blocks are matched sequentially according to the data frame length and the data is stored. Then, the address of the next free buffer block is matched sequentially according to the buffer block descriptor to complete the enqueueing of the entire frame.

7. The method for dynamic queue management of multi-priority services in time-sensitive network switching according to claim 1, characterized in that, The status information of the update queue linked list and shared cache includes the update queue frame count, queue head and tail pointers, and free cache block address cache occupancy and release information.

8. A method for dynamic queue management of multi-priority services in time-sensitive network switching according to claim 1, characterized in that, The strict priority scheduling principle prioritizes scheduling data frames from high-priority queues. The scheduling output information includes the output port, priority, and multicast flag information. Based on the scheduling information index, the corresponding dequeue queue is read, the queue head pointer is read, and the buffer block data is read sequentially until all data of a single data frame is read and transmitted through the port.

9. A method for dynamic queue management of multi-priority services in time-sensitive network switching according to claim 1, characterized in that, The specific process of the processing logic for multicast frames, which involves dequeuing in stages and finally updating the queue state uniformly, includes: During multicast frame scheduling, if some output ports are not idle, only the currently idle output ports are selected as the scheduling result to complete the dequeue transmission of some port data. After all multicast frames have been transmitted across all ports, release the storage space of all corresponding buffer blocks and update the free block address cache and the head pointer of the multicast queue.

10. A multi-priority service dynamic queue management system for time-sensitive network switching, characterized in that, include: The resource initialization module is used to initialize the queue linked list and shared cache resources, and build a queue storage architecture that adapts to multi-port and multi-priority services. The queue linked list is independently configured for different priority data frames of each destination port, and the shared cache resources are divided into standardized cache blocks and matched with corresponding cache management structures. The priority parsing module is used to receive data frames to be transmitted, extract the priority information and forwarding port information of the data frames, and map the priority information to the system's preset priority level. The queue enqueue module is used to match the corresponding queue list based on the mapped priority level and forwarding port information, perform data frame enqueue operation in combination with the queue idle state, and update the status information of the queue list and shared buffer. The queue dequeue module is used to perform dequeue scheduling on each queue linked list based on a strict priority scheduling principle, and read data frames from the queue linked list according to the scheduling result and complete port transmission. The resource reclamation module is used to release the corresponding shared buffer resources and update the status of idle buffers after the data frame transmission is completed. For multicast frames, it adopts a processing logic of dequeuing in stages and finally updating the queue status uniformly to complete the dynamic queue full-process management.

Citation Information

Patent Citations

  • Queue management method, device and system

    CN102308537A