Data transmission method, device and storage medium for multi-protocol hybrid network

Abstract

The present invention discloses a data transmission method, device and storage medium of a multi-protocol hybrid network, and relates to the field of network technology. The data transmission method of the multi-protocol hybrid network includes: parsing the acquired network traffic; when the protocol of the acquired network traffic is RoCE, determining the type of the network traffic as a low-latency type; when the protocol of the acquired network traffic is TCP and the network traffic meets the preset conditions, determining the type of the network traffic as a best-effort forwarding type; allocating the acquired network traffic to the corresponding queue; for the low-latency type queue and the best-effort forwarding type queue, respectively using the first queue length threshold and the second queue length threshold for congestion processing, the first queue length threshold is higher than the second queue length threshold; forwarding the network traffic in each queue. The present invention adopts a more aggressive congestion processing strategy to process TCP traffic that has a greater impact on RoCE traffic, thereby ensuring the performance of the RDMA network.

Description

Technical Field

[0001] The present invention relates to the field of network technology, and in particular to a data transmission method, device and storage medium for a multi-protocol hybrid network. Background Art

[0002] In application scenarios such as cloud data centers and edge computing, for high-performance computing services such as artificial intelligence (AI) services, the network needs to provide ultra-low latency, zero packet loss and other capabilities to avoid the network becoming a bottleneck. The deployment of such services usually adopts lossless network technology such as Remote Direct Memory Access (RDMA) to meet its requirements for ultra-low latency network characteristics. RDMA is based on the RoCE (RDMA over Converged Ethernet) protocol, which can retrieve data from the sender's address space and directly transmit it to the receiver's address space, thereby quickly completing the rapid exchange of data between computing node memories without the participation of kernel memory, greatly reducing the processing latency on the server side. Ethernet-based RoCE technology has gradually replaced special technologies such as Infiniband (IB) and become a mainstream technology. The latest RoCEv2 version uses IP (Internet Protocol) / UDP (User Datagram Protocol) to replace the IB network layer, providing IP routing and equal-cost multi-path routing (ECMP) capabilities, becoming the main deployment protocol for high-performance data center networks. Summary of the invention

[0003] After analysis, the inventors found that in scenarios such as edge data centers (DCs), limited by the size and cost of the network, edge networks for high-performance computing need to be compatible with existing traditional TCP (Transmission Control Protocol) network services while meeting RoCE network transmission. When the edge network simultaneously transmits mixed data of TCP traffic and RDMA traffic, the TCP congestion control algorithm will aggravate the congestion of RoCE and greatly reduce the performance of the RDMA network.

[0004] In the related art, TCP and RDMA traffic are mostly configured to different ports for independent data transmission. This solution is costly and resource configuration is inflexible.

[0005] A technical problem to be solved by the embodiments of the present invention is: how to provide a solution that reduces costs and guarantees RDMA network performance in a network that is compatible with RDMA and TCP network services.

[0006] According to a first aspect of some embodiments of the present invention, a data transmission method for a multi-protocol hybrid network is provided, including: parsing the acquired network traffic; when the protocol of the acquired network traffic is RoCE, determining the type of the network traffic as a low-latency type; when the protocol of the acquired network traffic is TCP and the network traffic meets a preset condition, determining the type of the network traffic as a best-effort forwarding type; allocating the acquired network traffic to a queue corresponding to the type of network traffic; for the low-latency type queue and the best-effort forwarding type queue, respectively using a first queue length threshold and a second queue length threshold for congestion processing, wherein the first queue length threshold is higher than the second queue length threshold; and forwarding the network traffic in each queue.

[0007] In some embodiments, for each of the low-latency type queue and the best-effort forwarding type queue, when the queue length is greater than the corresponding queue length threshold, the packets of the network traffic in the queue are marked, wherein the mark is used to instruct the sender of the traffic to reduce the sending rate.

[0008] In some embodiments, when the queue length is greater than the corresponding queue length threshold, the packets that need to be marked in the queue are determined based on the marking probability corresponding to the queue and using the WRED algorithm, wherein the marking probability of the first queue is lower than the marking probability of the second queue.

[0009] In some embodiments, using the WRED algorithm to determine the packets that need to be marked in the queue includes: in response to allocating packets of network traffic to the queue, using the WRED algorithm to determine the marking probability of the obtained packets; in response to the marking probability being greater than the marking probability corresponding to the queue, marking the packets to be sent in the queue.

[0010] In some embodiments, the mark is a congestion occurrence mark in the ECN mark.

[0011] In some embodiments, the preset condition is that the message length of the network traffic is greater than a preset value, and the life cycle length is greater than a preset length.

[0012] In some embodiments, the data transmission further includes: when the protocol of the acquired network traffic is TCP and the network traffic does not meet a preset condition, determining the type of the network traffic as a bandwidth guaranteed type.

[0013] In some embodiments, parsing the acquired network traffic includes: determining the protocol type of the network traffic based on one or more of the IP protocol number, TCP or UDP protocol identifier of the network traffic message; when the protocol type of the network traffic is the TCP protocol, determining the message length and life cycle of the network traffic.

[0014] According to a second aspect of some embodiments of the present invention, there is provided a data transmission device for a multi-protocol hybrid network, comprising: a parsing module, configured to parse the acquired network traffic; when the protocol of the acquired network traffic is RoCE, determining the type of the network traffic as a low-latency type; when the protocol of the acquired network traffic is TCP and the network traffic meets a preset condition, determining the type of the network traffic as a best-effort forwarding type; a traffic distribution module, configured to distribute the acquired network traffic to a queue corresponding to the type of network traffic; a congestion handling module, configured to perform congestion handling using a first queue length threshold and a second queue length threshold for a low-latency type queue and a best-effort forwarding type queue, respectively, wherein the first queue length threshold is higher than the second queue length threshold; and a forwarding module, configured to forward the network traffic in each queue.

[0015] According to a third aspect of some embodiments of the present invention, there is provided a data transmission device for a multi-protocol hybrid network, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute any one of the aforementioned data transmission methods for the multi-protocol hybrid network based on instructions stored in the memory.

[0016] According to a fourth aspect of some embodiments of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements any of the aforementioned methods for transmitting data in a multi-protocol hybrid network.

[0017] Some embodiments of the above invention have the following advantages or beneficial effects: when acquiring network traffic, RoCE traffic and TCP traffic that will have a greater impact on RoCE traffic are allocated to different queues according to the type of network traffic, and a more aggressive congestion handling strategy is adopted for TCP traffic that will have a greater impact on RoCE traffic. Thus, in a network compatible with RDMA and TCP network services, RDMA network performance can be guaranteed at a lower cost.

[0018] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0020] Figure 1 A schematic flow chart of a data transmission method in a multi-protocol hybrid network according to some embodiments of the present invention is shown.

[0021] Figure 2 A schematic structural diagram of a data transmission device in a multi-protocol hybrid network according to some embodiments of the present invention is shown.

[0022] Figure 3 A schematic structural diagram of a data transmission device for a multi-protocol hybrid network according to some other embodiments of the present invention is shown.

[0023] Figure 4 A schematic structural diagram of a data transmission device for a multi-protocol hybrid network according to some further embodiments of the present invention is shown. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] The relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

[0026] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0027] Technologies, methods, and apparatus known to ordinary technicians in the relevant field may not be discussed in detail, but where appropriate, such technologies, methods, and apparatus should be considered part of the authorization specification.

[0028] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0029] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0030] After further analysis, the inventors found that the performance of the RDMA network is very sensitive to latency and packet loss, especially packet loss caused by network congestion, which directly affects the performance of the RDMA network. When the edge network simultaneously transmits mixed data of TCP traffic and RDMA traffic, the TCP congestion control algorithm will cause too much queue cache to be occupied, which will impact the queue cache of RoCE, aggravate the congestion of RoCE, and greatly reduce the performance of the RDMA network.

[0031] Therefore, the inventors proposed a multi-queue scheduling method, which allocates some TCP traffic and RoCE traffic that will have a greater impact on RoCE traffic to different queues, and adopts different congestion handling mechanisms to ensure the performance of the RDMA network. Figure 1 An embodiment of a data transmission method for a multi-protocol hybrid network of the present invention is described.

[0032] Figure 1 FIG. 1 is a flow chart showing a method for transmitting data in a multi-protocol hybrid network according to some embodiments of the present invention. Figure 1 As shown, the data transmission method of the multi-protocol hybrid network of this embodiment includes steps S102 to S112.

[0033] In step S102, the acquired network traffic is parsed.

[0034] In some embodiments, the network traffic message is parsed, and the objects of the parsing include one or more of the IP source address, IP destination address, IP protocol number, TCP or UDP protocol identifier of the message.

[0035] In some embodiments, the protocol type of the network traffic is determined based on one or more of the IP protocol number, TCP or UDP protocol identifier of the network traffic message; when the protocol type of the network traffic is the TCP protocol, the message length and life cycle of the network traffic are determined.

[0036] For example, for the TCP protocol, there is an 8-bit protocol number in the IP header of the message, which is used to indicate the upper layer protocol of IP, where protocol number 6 represents the TCP protocol.

[0037] For example, the message structure of the RoCE v1 protocol adds a Layer 2 Ethernet message header to the original IB architecture message, and identifies the RoCE message through Ethertype 0x8915; the RoCE v2 protocol is based on the UDP / IP protocol. It adds a UDP header, IP header, and Layer 2 Ethernet message header to the original IB architecture message, and identifies the RoCE message through the UDP destination port number 4791.

[0038] In some embodiments, the parsed content may also include other information, such as the IP source address and IP destination address of the message including the network traffic. By combining the IP source address and the IP destination address and other parsed content, the user or service to which the traffic belongs can be determined. When the user or service has special requirements for QoS, such traffic can be divided into a separate queue, and a dedicated queue length threshold can be configured to perform congestion processing on the traffic in the queue differently from other queues.

[0039] In step S104, when the protocol of the acquired network traffic is RoCE, the type of the network traffic is determined to be a low-latency type.

[0040] In step S106, when the protocol of the acquired network traffic is TCP and the network traffic meets a preset condition, the type of the network traffic is determined to be a best-effort forwarding type.

[0041] In some embodiments, the TCP traffic that meets the preset condition refers to the TCP traffic that has a probability of causing congestion higher than a preset level. The probability of causing congestion can be measured, for example, by the length of the message and the life cycle.

[0042] In some embodiments, TCP traffic can be divided into "elephant flow" and "mouse flow". "Elephant flow" refers to traffic whose message length of network traffic is greater than a preset value and whose life cycle length is greater than a preset length, that is, traffic that performs large-scale, continuous data transmission; "mouse flow" refers to traffic whose message length is not greater than a preset value and whose life cycle length is not greater than a preset length, that is, traffic that performs small-scale, short-term data transmission. In some embodiments, TCP elephant flow is determined as a best-effort forwarding type because such traffic will seriously affect the performance of RoCE traffic after congestion occurs.

[0043] In some embodiments, when the protocol of the acquired network traffic is TCP and the network traffic does not meet the preset conditions, the type of network traffic is determined as a bandwidth guarantee type. For example, for a "TCP mouse flow", it is determined as a bandwidth guarantee type. Thus, for a network where TCP and RoCE traffic run together, the traffic can be allocated to three queues.

[0044] In step S108, the acquired network traffic is distributed to a queue corresponding to the type of the network traffic.

[0045] The performance of the RDMA network is very sensitive to latency and packet loss, so the RoCE protocol traffic is assigned to the low-latency type queue. TCP traffic that meets the preset conditions (such as TCP elephant flow) is more likely to impact the RDMA traffic, so it is assigned to the best-effort forwarding type queue.

[0046] In step S110, for a low-latency queue and a best-effort queue, congestion processing is performed using a first queue length threshold and a second queue length threshold, respectively, wherein the first queue length threshold is higher than the second queue length threshold.

[0047] When the queue length exceeds the corresponding length threshold, congestion handling is required. The length threshold corresponding to the low-latency type queue is smaller than the length threshold corresponding to the best-effort forwarding type queue, indicating that eligible TCP traffic needs to start congestion control earlier to reduce the impact on RoCE traffic.

[0048] In some embodiments, for bandwidth guarantee type queues, no queue length threshold for congestion handling is set. Since low latency type queues and best effort type queues have adopted corresponding congestion handling strategies, they will not affect bandwidth guarantee type queues.

[0049] In step S112, the network traffic in each queue is forwarded, for example, the next hop editing of the message and the egress forwarding operation are completed.

[0050] When acquiring network traffic, the method of the above embodiment allocates RoCE traffic and TCP traffic that has a greater impact on RoCE traffic to different queues according to the type of network traffic, and adopts a more aggressive congestion handling strategy for TCP traffic that has a greater impact on RoCE traffic. Thus, in a network compatible with RDMA and TCP network services, RDMA network performance can be guaranteed at a lower cost.

[0051] In some embodiments, for each of the low-latency type queue and the best-effort forwarding type queue, when the queue length is greater than the corresponding queue length threshold, the traffic of the network traffic in the queue that exceeds the queue length threshold is marked, wherein the mark is used to instruct the sender of the traffic to reduce the sending rate. Since the queue length threshold of the queue including the preset type of TCP traffic (e.g., TCP elephant flow) is smaller, the source server can be notified as early as possible to reduce the speed of this type of TCP traffic in advance.

[0052] In some embodiments, when the queue length is greater than the corresponding queue length threshold, the packets that need to be marked in the queue are determined by using a weighted random early detection (WRED) algorithm based on the marking probability corresponding to the queue, wherein the marking probability of the first queue is lower than the marking probability of the second queue.

[0053] When congestion occurs, the WRED algorithm generally selectively discards lower priority messages. However, since discarding messages will also increase the burden on the network, in some embodiments, the messages that need to be discarded are determined by the WRED algorithm as messages that need to be marked. For example, when a message is obtained, the WRED algorithm is used to calculate its "discard probability" as the marking probability for message marking in the embodiment of the present invention. If the calculated marking probability is greater than the preset marking probability corresponding to the corresponding queue, it is considered that the message needs to be marked.

[0054] Since the marking probability of the first queue is lower than that of the second queue, when the queue length of TCP traffic of a preset type such as TCP elephant flow is the same as that of RoCE traffic, TCP elephant flow has a higher probability of being marked, so that a greater degree of speed reduction can be performed. Therefore, the forwarding bandwidth of RoCE traffic is improved during congestion.

[0055] In some embodiments, in response to assigning a message of network traffic to a queue, a WRED algorithm is used to determine the marking probability of the obtained message; in response to the recorded probability being greater than the marking probability corresponding to the queue, the message to be sent in the queue is marked. When it is calculated that a certain obtained message needs to be marked, since the message needs to be added to the end of the queue, it is necessary to wait for a period of time before the mark can be transmitted to the sending end. Through the method of this embodiment, when it is calculated that there is a need to mark the message, the message to be sent is preferentially selected for marking instead of marking the message just obtained, so that the source server can be notified as soon as possible to slow down, which can more quickly and effectively relieve network pressure.

[0056] In some embodiments, the mark is a congestion occurrence mark in the ECN mark. The ECN mark detects path congestion through the cooperation of the sender, the receiver, and the router between the two, and actively reduces the sending rate of the sender, thereby avoiding packet loss caused by network congestion as early as possible.

[0057] In some embodiments, the message of the network traffic includes an ECN field, which has four values. When the ECN field value is 00, it means that the message does not support ECN. When congestion occurs, such messages are dropped based on the WRED algorithm; when the ECN field value is 01 or 10, when congestion occurs, its value is modified to 11 to indicate that the message has passed the congestion. When the receiving end receives a message with an ECN field value of 11, i.e., a congestion occurrence mark, it sends a confirmation message including a speed reduction indication to the sending end. As a result, the sending end reduces the sending rate to relieve network pressure and improves the forwarding bandwidth of RoCE traffic during congestion.

[0058] Reference below Figure 2 An embodiment of a data transmission device for a multi-protocol hybrid network of the present invention is described.

[0059] Figure 2 FIG. 1 is a schematic diagram showing a structure of a data transmission device for a multi-protocol hybrid network according to some embodiments of the present invention. Figure 2 As shown, the data transmission device 20 of the multi-protocol hybrid network of this embodiment includes: a parsing module 210, configured to parse the acquired network traffic; when the protocol of the acquired network traffic is RoCE, the type of the network traffic is determined to be a low-latency type; when the protocol of the acquired network traffic is TCP and the network traffic meets the preset conditions, the type of the network traffic is determined to be a best-effort forwarding type; a traffic distribution module 220, configured to distribute the acquired network traffic to the queue corresponding to the type of network traffic; a congestion processing module 230, configured to use a first queue length threshold and a second queue length threshold for congestion processing for a low-latency type queue and a best-effort forwarding type queue, respectively, wherein the first queue length threshold is higher than the second queue length threshold; a forwarding module 240, configured to forward the network traffic in each queue.

[0060] In some embodiments, the congestion handling module 230 is further configured to mark the packets of the network traffic in each of the low-latency type queues and the best-effort forwarding type queues when the queue length is greater than the corresponding queue length threshold, wherein the mark is used to instruct the sender of the traffic to reduce the sending rate.

[0061] In some embodiments, the congestion processing module 230 is further configured to determine the packets that need to be marked in the queue by using the WRED algorithm based on the marking probability corresponding to the queue when the queue length is greater than the corresponding queue length threshold, wherein the marking probability of the first queue is lower than the marking probability of the second queue.

[0062] In some embodiments, the congestion handling module 230 is further configured to determine the marking probability of the obtained message using the WRED algorithm in response to allocating the message of the network traffic to the queue; and mark the message to be sent in the queue in response to the marking probability being greater than the marking probability corresponding to the queue.

[0063] In some embodiments, the mark is a congestion occurrence mark in the ECN mark.

[0064] In some embodiments, the preset condition is that the message length of the network traffic is greater than a preset value, and the life cycle length is greater than a preset length.

[0065] In some embodiments, the parsing module 210 is further configured to determine the type of network traffic as a bandwidth guaranteed type when the protocol of the acquired network traffic is TCP and the network traffic does not meet a preset condition.

[0066] In some embodiments, the parsing module 210 is further configured to determine the protocol type of the network traffic based on one or more of the IP source address, IP destination address, IP protocol number, TCP or UDP protocol identifier of the network traffic message; when the protocol type of the network traffic is TCP protocol, determine the message length and life cycle of the network traffic.

[0067] Figure 3 FIG. 2 shows a schematic diagram of a structure of a data transmission device for a multi-protocol hybrid network according to some other embodiments of the present invention. Figure 3 As shown, the data transmission device 30 of the multi-protocol hybrid network of this embodiment includes: a memory 310 and a processor 320 coupled to the memory 310, and the processor 320 is configured to execute the data transmission method of the multi-protocol hybrid network in any of the aforementioned embodiments based on the instructions stored in the memory 310.

[0068] The memory 310 may include, for example, a system memory, a fixed non-volatile storage medium, etc. The system memory may store, for example, an operating system, an application program, a boot loader, and other programs.

[0069] Figure 4 FIG. 2 shows a schematic diagram of a structure of a data transmission device for a multi-protocol hybrid network according to some other embodiments of the present invention. Figure 4As shown, the data transmission device 40 of the multi-protocol hybrid network of this embodiment includes: a memory 410 and a processor 420, and may also include an input / output interface 430, a network interface 440, a storage interface 450, etc. These interfaces 430, 440, 450 and the memory 410 and the processor 420 may be connected, for example, via a bus 460. Among them, the input / output interface 430 provides a connection interface for input / output devices such as a display, a mouse, a keyboard, and a touch screen. The network interface 440 provides a connection interface for various networked devices. The storage interface 450 provides a connection interface for external storage devices such as SD cards and USB flash drives.

[0070] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements any of the aforementioned methods for transmitting data in a multi-protocol hybrid network.

[0071] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable non-transient storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0072] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0073] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0074] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0075] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A data transmission method for a multi-protocol hybrid network, comprising: Parse the acquired network traffic; When the protocol of the acquired network traffic is Remote Direct Data Access (RoCE) over Converged Ethernet, determining the type of the network traffic as a low-latency type; When the protocol of the acquired network traffic is the Transmission Control Protocol TCP and the network traffic meets a preset condition, determining the type of the network traffic as a best-effort forwarding type, wherein the preset condition is that the message length of the network traffic is greater than a preset value and the life cycle length is greater than a preset length; When the protocol of the acquired network traffic is TCP and the network traffic does not meet the preset condition, determining the type of the network traffic as a bandwidth guaranteed type; Allocating the acquired network traffic to a queue corresponding to the type of the network traffic; For a low-latency queue and a best-effort forwarding queue, a first queue length threshold and a second queue length threshold are respectively used to perform congestion processing; for a bandwidth-guaranteed queue, no queue length threshold for congestion processing is set, wherein the first queue length threshold is higher than the second queue length threshold; for each of the low-latency queue and the best-effort forwarding queue, when the queue length is greater than the corresponding queue length threshold, a message of the network traffic in the queue is marked, wherein the mark is used to instruct a sender of the traffic to reduce a sending rate; Forwards network traffic in each queue.

2. The data transmission method according to claim 1, wherein: When the queue length is greater than the corresponding queue length threshold, the packets that need to be marked in the queue are determined by using a weighted random early detection (WRED) algorithm based on the marking probability corresponding to the queue, wherein the marking probability of the first queue is lower than the marking probability of the second queue.

3. The data transmission method according to claim 2, wherein: The using of the WRED algorithm to determine the packets in the queue that need to be marked includes: In response to allocating the packets of the network traffic to the queue, determining the marking probability of the obtained packets using a WRED algorithm; In response to the marking probability being greater than the marking probability corresponding to the queue, marking the messages to be sent in the queue.

4. The data transmission method according to any one of claims 1 to 3, wherein: The mark is a congestion occurrence mark in the ECN mark.

5. The data transmission method according to claim 1, wherein: The network traffic obtained by the analysis includes: Determine the protocol type of the network traffic according to one or more of the network protocol IP protocol number, TCP or User Datagram Protocol UDP identifier of the message of the network traffic; When the protocol type of the network traffic is the TCP protocol, determine the message length and life cycle of the network traffic.

6. A data transmission device for a multi-protocol hybrid network, comprising: a parsing module configured to parse the acquired network traffic; When the protocol of the acquired network traffic is RoCE, the type of the network traffic is determined to be a low-latency type; when the protocol of the acquired network traffic is TCP and the network traffic meets a preset condition, the type of the network traffic is determined to be a best-effort forwarding type, and the preset condition is that the message length of the network traffic is greater than a preset value and the life cycle length is greater than a preset length; When the protocol of the acquired network traffic is TCP and the network traffic does not meet the preset condition, determining the type of the network traffic as a bandwidth guaranteed type; A traffic distribution module, configured to distribute the acquired network traffic to a queue corresponding to the type of the network traffic; a congestion processing module configured to perform congestion processing using a first queue length threshold and a second queue length threshold for a low-latency queue and a best-effort forwarding queue, respectively, and not setting a queue length threshold for congestion processing for a bandwidth-guaranteed queue, wherein the first queue length threshold is higher than the second queue length threshold, and for each of the low-latency queue and the best-effort forwarding queue, when the queue length is greater than the corresponding queue length threshold, marking a message of network traffic in the queue, wherein the mark is used to instruct a sender of the traffic to reduce a sending rate; The forwarding module is configured to forward network traffic in each queue.

7. A data transmission device for a multi-protocol hybrid network, comprising: Memory; as well as A processor coupled to the memory, the processor being configured to execute the data transmission method for a multi-protocol hybrid network according to any one of claims 1 to 5 based on instructions stored in the memory.

8. A computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the data transmission method for a multi-protocol hybrid network according to any one of claims 1 to 5 is implemented.

Images