An embedded-based intelligent congestion management network communication device

Abstract

The application belongs to the technical field of embedded computer network communication, and particularly relates to an intelligent congestion management network communication device based on embedding, which comprises an SOC chip with an AI component, an optoelectronic transceiver, a gigabit Ethernet card and an MCU; the size flow proportion in current network flow, queue cache occupancy rate and flow throughput condition characteristics are extracted and transmitted to the AI component in real time for data analysis, and the ECN threshold value is dynamically adjusted in real time through a flow model; the MAC layer sends a congestion release notification based on the threshold, and restores the sending rate of the sending end; when the ECN and PFC are invalid at the same time, network paralysis occurs, the fault is recorded and the device is restarted, and meanwhile, the remaining hardware faults can also be recorded and reported on the bus. The application can intelligently adjust the ECN value in real time, and effectively relieve the congestion of Ethernet data flow.

Description

Technical Field

[0001] This invention belongs to the field of embedded computer network communication technology, specifically relating to an embedded intelligent congestion management network communication device. Background Technology

[0002] To adapt to the complex and ever-changing aerial environment and meet the requirements of modern digital transformation, the scale of storage networks is increasing daily, and the amount of data that needs to be processed is growing larger and larger. RDMA (Remote Direct Memory Access) has become the mainstream technology for storage networks.

[0003] Lossless Ethernet refers to the enhancement of the Ethernet protocol used to achieve high-performance data transmission. It achieves flow control through traffic allocation and eliminates packet loss due to queue overflow, thereby achieving the goal of zero packet loss, low latency, and high throughput in the network.

[0004] The PFC function performs flow control on packets based on dot1p priority. An Ethernet link is divided into 8 virtual channels, each with its own priority level and corresponding to one of 8 interface buffers. Each virtual channel can be paused and restarted independently.

[0005] ECN is a congestion notification technology. The ECN function uses the DS field in the IP packet header to mark whether the packet transmission path has experienced congestion. Terminal devices that support this function can adjust the packet sending rate according to the ECN marking status of the packet, thereby alleviating congestion.

[0006] In the RFC2481 standard, the DS and the last two bits in the IP header are defined as the ECN field.

[0007] Bit 6 is used to identify whether the sending device supports the ECN function;

[0008] Bit 7 is used to indicate whether the message has experienced congestion on the transmission path;

[0009] When the ECN field is 00, it indicates that the message does not support the ECN function.

[0010] When the ECN field is 01, it indicates that the message supports the ECN function;

[0011] When the ECN field is 11, it indicates that the packet experienced congestion in the forwarding path;

[0012] Existing Ethernet data streams are frequently congested. Summary of the Invention

[0013] In view of this, in order to achieve intelligent congestion management and meet the requirements of large data volume communication and low latency transmission, this application provides an embedded intelligent congestion management network communication device that can intelligently adjust the ECN value in real time to effectively alleviate Ethernet data stream congestion.

[0014] The technical solution of this invention is as follows:

[0015] An embedded intelligent congestion management network communication device includes an SOC chip with AI components, an optoelectronic transceiver, a 10 Gigabit Ethernet card, and an MCU, wherein:

[0016] Ethernet traffic data is converted by the optoelectronic transceiver via the Ethernet Fibre Channel protocol and then sent to the SOC chip with an AI component. The SOC chip is equipped with DDR4, QSPI FLASH, 1TB SSD, and NVRAM. The PS end of the SOC chip extracts the proportion of large and small flows, queue buffer occupancy rate, and traffic throughput characteristics in the current network traffic and transmits them to the AI ​​component in real time for data analysis. The AI ​​component dynamically adjusts the ECN threshold value in real time through the traffic model to alleviate network congestion.

[0017] The embedded intelligent congestion management network device uses the MCU as a health management device. When ECN and PFC fail simultaneously and network paralysis occurs, the PL terminal of the SOC chip records the fault in NVRAM and sends the fault information to the MCU via the IIC bus. The MCU then reports the fault to the end system via the CAN bus and resets the SOC chip.

[0018] After network congestion management is completed, the PL terminal of the SOC chip further performs Ethernet traffic data forwarding processing; the processing result is sent to the next-level communication device after conversion by the optoelectronic transceiver.

[0019] Furthermore, the optoelectronic transceiver transmits the Ethernet data stream to the processor via the PCIe bus, and the PS terminal of the SOC chip extracts the characteristics of the current network stream, including the proportion of large and small streams, queue buffer occupancy rate, and traffic throughput.

[0020] Furthermore, the PL terminal of the SOC chip reads the network flow feature information extracted by the PS terminal in real time and determines whether it conforms to the traffic model. If it conforms, the existing model is used to infer the ECN threshold. If it does not conform to the traffic model, the algorithm is called to obtain the real-time global optimal threshold.

[0021] Furthermore, the DDR4 is used to cache Ethernet stream data that is to be processed and that has already been processed.

[0022] Furthermore, the QSPI FLASH is used to store the BIT test program, BOOT boot program, operating system, and application programs.

[0023] Furthermore, the NVRAM is used to record fault information.

[0024] Furthermore, the SSD is used to store the trained Ethernet traffic model and algorithm model data.

[0025] Furthermore, the MCU is used for health monitoring and management, adopts an independent power supply channel, and monitors the voltage status of each chip in the embedded intelligent congestion management network communication device in real time. If a problem occurs, the fault status is recorded and the fault is reported through the CAN bus. When ECN and PFC fail at the same time, the MCU reports the fault status and resets the SOC chip.

[0026] Furthermore, the embedded intelligent congestion management network communication device is equipped with a 10 Gigabit Ethernet card to support electrical signal interface communication; the processing result of the SOC chip is sent to the lower-level communication device after conversion by the optoelectronic transceiver, or communicates with the host computer after conversion by the 10 Gigabit Ethernet card.

[0027] Furthermore, the MCU also reports other recorded hardware faults via the CAN bus.

[0028] The advantages of this invention are:

[0029] 1. When the proportion of small flows in the queue is high, dynamically lower the ECN trigger threshold to ensure the low latency requirements of most small flows; when the proportion of large flows in the queue is high, dynamically raise the ECN trigger threshold to ensure the high throughput requirements of most large flows.

[0030] 2. Employing AI components utilizes Ethernet traffic models and algorithm models to evaluate effective ECN thresholds, enabling real-time adjustment of ECN threshold values. This effectively alleviates congestion while reducing the impact on network transmission efficiency, demonstrating a high degree of intelligence and fast processing speed.

[0031] 3. The device has a health monitoring and reporting function, which can detect the functional status of each component in the device, record the Ethernet congestion management status, and participate in network congestion management when necessary, which can significantly shorten the fault location time. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 A schematic diagram of the structure of an embedded-based intelligent congestion management network communication device provided in this application;

[0034] Figure 2 A schematic diagram of the ECN field in the IPv4 packet header provided in this application;

[0035] Figure 3 A flowchart illustrating the system operation provided for this application;

[0036] Figure 4 The flowchart provided for this application. Detailed Implementation

[0037] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0038] The following specific examples illustrate the implementation of this disclosure. Those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0039] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using other structures and / or functionalities besides one or more of the aspects set forth herein.

[0040] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this disclosure. The drawings only show the components related to this disclosure and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0041] Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0042] In one embodiment of the present invention, an embedded intelligent congestion management network communication device is proposed, see reference. Figures 1-4 The embedded intelligent congestion management network device includes an SOC chip with AI components, an optoelectronic transceiver, a 10 Gigabit Ethernet card, and an MCU, wherein:

[0043] Ethernet traffic data is converted by the optoelectronic transceiver via the Ethernet Fibre Channel protocol and then sent to the SOC chip with an AI component. The SOC chip is equipped with DDR4, QSPI FLASH, 1TB SSD, and NVRAM. The PS end of the SOC chip extracts the proportion of large and small flows, queue buffer occupancy rate, and traffic throughput characteristics in the current network traffic and transmits them to the AI ​​component in real time for data analysis. The AI ​​component dynamically adjusts the ECN threshold value in real time through the traffic model to alleviate network congestion.

[0044] The embedded intelligent congestion management network device uses the MCU as a health management device. When ECN and PFC fail simultaneously and network paralysis occurs, the PL terminal of the SOC chip records the fault in NVRAM and sends the fault information to the MCU via the IIC bus. The MCU then reports the fault to the end system via the CAN bus and resets the SOC chip.

[0045] After network congestion management is completed, the PL terminal of the SOC chip further performs Ethernet traffic data forwarding processing; the processing result is sent to the next-level communication device after conversion by the optoelectronic transceiver.

[0046] In this embodiment, the optoelectronic transceiver transmits the Ethernet data stream to the processor via the PCIe bus, and the PS terminal of the SOC chip extracts the characteristics of the current network stream, including the proportion of large and small streams, the queue buffer occupancy rate, and the traffic throughput.

[0047] In this embodiment, the PL terminal of the SOC chip reads the network flow feature information extracted by the PS terminal in real time and determines whether it conforms to the traffic model. If it conforms, the existing model is used to infer the ECN threshold. If it does not conform to the traffic model, the algorithm is called to obtain the real-time global optimal threshold.

[0048] In this embodiment, the DDR4 is used to cache Ethernet stream data that is to be processed and that has already been processed.

[0049] In this embodiment, the QSPI FLASH is used to store the BIT test program, BOOT bootloader, operating system, and application programs.

[0050] In this embodiment, the NVRAM is used to record fault information.

[0051] In this embodiment, the SSD is used to store the trained Ethernet traffic model and algorithm model data.

[0052] In this embodiment, the MCU is used for health monitoring and management. It adopts an independent power supply channel to monitor the voltage status of each chip in the embedded intelligent congestion management network communication device in real time. If a problem occurs, the fault status is recorded and the fault is reported through the CAN bus. When ECN and PFC fail at the same time, the MCU reports the fault status and resets the SOC chip.

[0053] In this embodiment, the embedded intelligent congestion management network communication device is equipped with a 10 Gigabit Ethernet card to support electrical signal interface communication; the processing result of the SOC chip is sent to the lower-level communication device after conversion by the optoelectronic transceiver, or communicates with the host computer after conversion by the 10 Gigabit Ethernet card.

[0054] In this embodiment, the MCU also reports other recorded hardware faults via the CAN bus.

[0055] This includes an optoelectronic transceiver, a SoC chip with integrated AI components, an MCU, and a 10 Gigabit Ethernet card, among which:

[0056] In this embodiment, an embedded intelligent congestion management network communication device uses an optoelectronic transceiver to convert Ethernet data streams into media and send them to the PS terminal of the SOC chip. The PS terminal extracts the characteristics of the current network flow, such as the proportion of large and small flows, queue buffer occupancy rate, and traffic throughput, and caches this data in the PS terminal's DDR. The SOC's PL terminal reads the extracted network flow feature information from the PS terminal's DDR in real time into the PL terminal's DDR via DMA. The AI ​​component extracts the network flow feature information from the PL terminal's DDR to perform AI calculations.

[0057] The AI ​​component first determines whether the network flow feature information conforms to the traffic model. If it does, it uses the existing model to infer the ECN threshold. If it does not conform to the traffic model, it calls the algorithm to obtain the real-time global optimal threshold.

[0058] At the MAC layer, when a packet entering the queue is detected and the queue length exceeds the ECN low threshold, the ECN field value of the packet is modified using linear drop probability and forwarded. When a packet with an ECN field value of 11 is detected, the receiver acts as the congestion notification initiator, sending a congestion notification to the packet initiator at a fixed interval T1, requesting the sender to reduce its packet transmission rate by 50%. When no more packets with an ECN field value of 11 are received at interval T2, a congestion relief notification is sent, and the sender's transmission rate is restored. If congestion notifications are still received at interval T3, the sender's packet transmission rate is further reduced by 50%. If congestion notifications are still received at interval T4, it is considered that reducing the sender's packet transmission rate is insufficient to resolve the current network congestion, and PFC synchronous intervention is required.

[0059] When network congestion is severe and PFC intervention is required, the PS end of the SOC monitors the queue buffer consumption on each channel. When the consumption exceeds the buffer threshold at interval T4, the receiver sends a backpressure signal (PFC PAUSE frame) to the upstream device. Upon receiving the backpressure signal, the upstream device stops sending packets of the corresponding priority and stores the data in the local interface's buffer space. Simultaneously, the sender starts a deadlock detection timer, checking for PFC PAUSE frames received by the priority queue during the detection period at interval T5. If the queue remains in the PFC backpressure trigger threshold state (i.e., continuously receiving PFC PAUSE frames during the detection period), it is considered to be in a deadlock state. When a deadlock is detected on a certain interface, an automatic recovery timer is started. During the automatic recovery period, the PFC function and PFC deadlock detection function are disabled, and received PFC PAUSE frames are ignored. When the automatic recovery timer expires at interval T6, the device activates the PFC function and PFC deadlock detection function. If PFC deadlock continues to occur, the SOC's PL terminal records the fault in NVRAM and sends the fault information to the MCU via the IIC bus. The MCU then reports the fault to the end system via the CAN bus and resets the SOC chip. When the queue buffer consumption returns to normal within the reverse voltage signal transmission interval T7, the transmission of reverse voltage signals to upstream devices stops.

[0060] After network congestion management is completed, the PL end of the SOC further performs Ethernet traffic data forwarding and other tasks. After conversion by an optoelectronic transceiver, the data is sent to the lower-level communication device. Alternatively, it can communicate with the host computer through a 10 Gigabit network card, thereby realizing an embedded intelligent congestion management network communication device.

[0061] Specifically, the optoelectronic transceiver transmits data to the SOC chip via the PCIe bus Ethernet data stream, and the AI ​​component uses a dedicated traffic model and algorithm to estimate the ECN value.

[0062] Specifically, the SOC chip is connected to the 10 Gigabit Ethernet adapter via the PCIe bus.

[0063] Specifically, the SOC chip connects to the MCU via the IIC bus.

[0064] Specifically, the DDR4 is used to cache Ethernet stream data that is to be processed and that has already been processed.

[0065] Specifically, the FLASH is used to store the BIT test program, BOOT boot program, operating system, and application programs.

[0066] Specifically, NVRAM is used to record fault information.

[0067] Specifically, the SSD is used to store the trained Ethernet traffic model and algorithm model data.

[0068] like Figure 1 As shown, this embodiment provides an embedded intelligent congestion management network communication device. Ethernet data streams are converted to media via an optoelectronic transceiver and sent to the PS (Power Supply) terminal of the SOC (System-on-a-Chip) chip. The PS terminal extracts features such as the proportion of large and small flows, queue buffer occupancy, and traffic throughput in the current network flow, and synchronously starts a feature extraction timer. When the timer expires and the traffic feature extraction flag register is still 1, the PL (Power Supply) terminal is enabled to participate in feature extraction, using a pipelined approach to process in parallel with the PS terminal. When the traffic feature extraction flag register is 0, it is considered that the feature extraction of the current traffic frame has been completed, and its data is cached in the PS terminal's DDR (Data Memory). The SOC's PL terminal reads the extracted network flow feature information from the PS terminal's DDR into the PL terminal's DDR in real time via DMA (Digital Messaging). The AI ​​component extracts the network flow feature information from the PL terminal's DDR for AI calculations.

[0069] The AI ​​component first determines whether the network flow feature information conforms to the traffic model. If it does, it uses the existing model to infer the ECN threshold. If it does not conform to the traffic model, it calls the algorithm to obtain the real-time global optimal threshold.

[0070] At the MAC layer, when a traffic packet passing through the device is detected to have entered the queue and the queue length exceeds the low threshold of ECN, the ECN field value of the packet is modified using linear drop probability and then forwarded.

[0071] The congestion management working mode is shown in Table 1.

[0072] Table 1. Congestion Management Working Mode

[0073]

[0074]

[0075] S1: When a message with an ECN field value of 11 is detected, the receiving end acts as the congestion notification initiating device and sends a congestion notification to the message initiating end at a fixed interval T1, requesting the sending end to reduce the message sending rate by 50%.

[0076] S2: When no more packets with ECN field 11 are received after interval T2, send a congestion relief notification and restore the sending rate of the sending end;

[0077] S3: If a congestion notification is still received at interval T3, further reduce the message sending rate of the sender by 50%;

[0078] S4: If congestion notifications are still received at interval T4, it is assumed that reducing the message sending rate at the sending end is insufficient to resolve the current network congestion, and PFC intervention is required. When network congestion is severe at interval T4 and PFC intervention is necessary, the PS end of the SOC monitors the queue buffer consumption on each channel. If the consumption exceeds the buffer threshold, the receiving end sends a backpressure signal (PFC PAUSE frame) to the upstream device. Upon receiving the backpressure signal, the upstream device stops sending messages of the corresponding priority and stores the data in the local interface's buffer space. Simultaneously, the sending end starts a deadlock detection timer.

[0079] S5: Within the detection period T5, detect the PFC PAUSE frames received by this priority queue. If the queue remains in the PFC backpressure trigger threshold state (i.e., continuously receives PFC PAUSE frames within the detection period), it is considered to be in a deadlock state. When a deadlock is detected on an interface, start the automatic recovery timer. During the automatic recovery period, disable the PFC function and the PFC deadlock detection function, and ignore the received PFC PAUSE frames.

[0080] S6: When the automatic recovery timer T6 expires, the device enables the PFC function and PFC deadlock detection function. If PFC deadlock continues to occur, the PL terminal of the SOC will record the fault in the NVRAM and send the fault information to the MCU via the IIC bus. The MCU will then report the fault to the end system via the CAN bus and reset the SOC chip.

[0081] S7: When the interval T7 for sending back pressure signals returns to normal and the queue buffer consumption returns to normal, stop sending back pressure signals to upstream devices.

[0082] After completing network congestion management, the PL end of the SOC further performs Ethernet traffic data forwarding and other tasks, then writes the forwarding flag high. After conversion by an optoelectronic transceiver, it is sent to the lower-level communication device. Alternatively, it can communicate with the host computer through a 10 Gigabit network card, thereby realizing an embedded intelligent congestion management network communication device.

[0083] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. An embedded-based intelligent congestion management network communication device, characterized in that, The embedded intelligent congestion management network device includes a SOC chip with AI components, an optoelectronic transceiver, a 10 Gigabit Ethernet card, and an MCU, wherein: Ethernet traffic data is converted by the optoelectronic transceiver via the Ethernet Fibre Channel protocol and then sent to the SOC chip with an AI component. The SOC chip is equipped with DDR4, QSPI FLASH, 1TB SSD, and NVRAM. The PS end of the SOC chip extracts the proportion of large and small flows, queue buffer occupancy rate, and traffic throughput characteristics in the current network traffic and transmits them to the AI ​​component in real time for data analysis. The AI ​​component dynamically adjusts the ECN threshold value in real time through the traffic model to alleviate network congestion. The embedded intelligent congestion management network device uses the MCU as a health management device. When ECN and PFC fail simultaneously and network paralysis occurs, the PL terminal of the SOC chip records the fault in NVRAM and sends the fault information to the MCU via the IIC bus. The MCU then reports the fault to the end system via the CAN bus and resets the SOC chip. After network congestion management is completed, the PL terminal of the SOC chip further performs Ethernet traffic data forwarding processing; the processing result is sent to the next-level communication device after conversion by the optoelectronic transceiver.

2. The embedded-based intelligent congestion management network communication device according to claim 1, characterized in that, The optoelectronic transceiver transmits Ethernet data streams to the processor via the PCIe bus, and the PS terminal of the SOC chip extracts the characteristics of the current network stream, including the proportion of large and small streams, queue buffer occupancy rate, and traffic throughput.

3. The embedded-based intelligent congestion management network communication device according to claim 2, characterized in that, The PL terminal of the SOC chip reads the network flow feature information extracted by the PS terminal in real time and determines whether it conforms to the traffic model. If it conforms, the ECN threshold is inferred using the existing model. If it does not conform to the traffic model, the algorithm is called to obtain the real-time global optimal threshold.

4. The embedded-based intelligent congestion management network communication device according to claim 3, characterized in that, The DDR4 is used to cache Ethernet stream data that is to be processed and that has already been processed.

5. The embedded-based intelligent congestion management network communication device according to claim 4, characterized in that, The QSPI FLASH is used to store the BIT test program, BOOT boot program, operating system, and application programs.

6. The embedded-based intelligent congestion management network communication device according to claim 5, characterized in that, The NVRAM is used to record fault information.

7. The embedded-based intelligent congestion management network communication device according to claim 6, characterized in that, The SSD is used to store the trained Ethernet traffic model and algorithm model data.

8. The embedded-based intelligent congestion management network communication device according to claim 7, characterized in that, The MCU is used for health monitoring and management. It adopts an independent power supply channel and monitors the voltage status of each chip in the embedded intelligent congestion management network communication device in real time. If a problem occurs, it records the fault and reports the fault through the CAN bus. When ECN and PFC fail at the same time, the MCU reports the fault status and resets the SOC chip.

9. The embedded-based intelligent congestion management network communication device according to claim 8, characterized in that, The embedded intelligent congestion management network communication device is equipped with a 10 Gigabit Ethernet card to support electrical signal interface communication; the processing result of the SOC chip is sent to the lower-level communication device after conversion by the optoelectronic transceiver, or communicates with the host computer after conversion by the 10 Gigabit Ethernet card.

10. The embedded-based intelligent congestion management network communication device according to claim 9, characterized in that, The MCU also reports other recorded hardware faults via the CAN bus.

Images