WO2022105884A1

WO2022105884A1 - Data transmission method and apparatus, network device, and storage medium

Info

Publication number: WO2022105884A1
Application number: PCT/CN2021/131823
Authority: WO
Inventors: 金浩; 屠要峰; 韩银俊; 郭斌; 许军宁; 杨洪章
Original assignee: 中兴通讯股份有限公司
Priority date: 2020-11-23
Filing date: 2021-11-19
Publication date: 2022-05-27
Also published as: CN113485823A

Abstract

Embodiments of the present application relate to the field of storage data transmission, and in particular to a data transmission method and apparatus, a network device, and a storage medium. In the embodiments of the present application, the data transmission method comprises: reading data from a source network node to a memory space by means of a source network card; forwarding information of the memory space to a target user mode protocol stack, and constructing a message of a target network protocol type by the target user mode protocol stack according to the information of the memory space, the message carrying a data address in the memory space; and sending the data in the memory space to a target network node by means of a target network card on the basis of the message of the target network protocol type.

Description

Data transmission method, device, network device, storage medium

cross reference

This application is based on the Chinese patent application with the application number "202011323780.5" and the filing date is November 23, 2020, and claims the priority of the Chinese patent application. The entire content of the Chinese patent application is hereby incorporated by way of reference. Application.

technical field

The embodiments of the present application relate to the field of storage data transmission, and in particular, to a data transmission method, apparatus, network device, and storage medium.

Background technique

When the source network transmits data to the target network, but the protocol stacks of the source network and the target network are not equal, the network device needs to process the transmitted data to meet the data requirements of the target network. It is transmitted to the network device, the data is copied from the user mode in the network device to the operating system protocol stack of the kernel mode, and the data is processed through the operating system protocol stack, so that the processed data satisfies the transmission protocol of the target network, and then the The processed data is sent to the target network.

However, during data processing, system state switching and data copying are required to copy data from the user state to the operating system protocol stack of the kernel state, resulting in a large system CPU resource overhead.

SUMMARY OF THE INVENTION

The embodiment of the present application provides a data transmission method, including: reading data from a source network node to a memory space through a source network card; forwarding information in the memory space to a target user mode protocol stack, and the target user mode protocol stack is based on the memory space. The space information constructs the message of the target network protocol type, the message carries the data address in the memory space, and the data in the memory space is sent to the target network node through the target network card based on the message of the target network protocol type.

An embodiment of the present application also provides a data transmission device, including: a data reading module, used for reading data from a source network node to a memory space through a source network card; a data forwarding module, used for forwarding information in the memory space to The target user mode protocol stack, the target user mode protocol stack constructs the message of the target network protocol type according to the information of the memory space, and the message carries the data address in the memory space; the data sending module is used to pass the target network card based on the target network protocol type. The message sends the data in the memory space to the target network node.

An embodiment of the present application further provides a network device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a program that can be executed by the at least one processor Instructions that are executed by the at least one processor to enable the at least one processor to perform the above-described data transfer method.

Embodiments of the present application further provide a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the above-mentioned data transmission method is implemented.

Description of drawings

1 is a schematic diagram of a data transmission network according to the related art;

2 is a flowchart of a data transmission method according to the first embodiment of the present application;

3 is a schematic diagram of a data transmission network according to the first embodiment of the present application;

4 is a flowchart of a data transmission method according to a second embodiment of the present application;

5 is a schematic diagram of a data transmission network according to a second embodiment of the present application;

6 is a schematic diagram of a data transmission apparatus according to a third embodiment of the present application;

FIG. 7 is a schematic structural diagram of a network device according to a fourth embodiment of the present application.

Detailed ways

In order to make the objectives, technical solutions and advantages of the embodiments of the present application more clear, each embodiment of the present application will be described in detail below with reference to the accompanying drawings. However, those of ordinary skill in the art can understand that, in each embodiment of the present application, many technical details are provided for the reader to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in the present application can be realized. The following division of each embodiment is for the convenience of description, and should not constitute any limitation to the specific implementation of the present application, and each embodiment may be combined with each other and referenced on the premise of not contradicting each other.

In the process of data transmission, there may be a situation where the network stacks of the transceiver and transmission are not equal. The related technology uses the operating system protocol stack to convert the protocol. During the conversion process, multiple data copies and system state switching are required. Resource overhead is high. The following takes the source network, that is, the sending end network is a Remote Direct Memory Access (RDMA) network, the target network, that is, the receiving end network is ordinary Ethernet, and the remote read operation of the client in the distributed storage scenario is used as an example to illustrate the description. The data transmission process of the related art. As shown in Figure 1, node1 and node2 are network nodes in RDMA, namely source network nodes, Ethernet network interface card (ENIC) is a common Ethernet network card, RNIC (a network adapter that can realize RDMA function, RDMA enable network interface controller) network card is an RDMA network card with an RDMA engine. The client connects to the master node of the distributed file system through the Transmission Control Protocol (TCP) link. The master node reads the required target data from other nodes through RDMA, writes into the local RDMA memory, may also need to be written to the local storage, and then return data to the client through the TCP link, the process is as follows;

Step 1: The master node of the storage system reads data from other nodes in the cluster to the MR (Memeory Regin) space of this node through the RNIC network card.

Step 2: The master node of the storage system copies the data in the MR to the socket, and then sends it to the buffer TxBuf. If necessary, it can be written to the local cache storage.

Step 3: Send the buffer data to the operating system protocol stack OS protocol stack. The data in the buffer TxBuf belongs to the application layer. To send it to the operating system protocol stack, it needs to copy the data and switch the user state to the kernel state.

Step 4: The operating system protocol stack sends the TCP packet to the network card queue.

Step 5: Finally, the ENIC network card sends the TCP packet to the network. Sent over the network to the target node.

The whole process described above requires system state switching, data copying, high system CPU resource overhead, and significant client response delay. This forwarding process becomes the bottleneck of the entire storage system.

Therefore, the first embodiment of the present application proposes a data transmission method, which can be applied to network devices such as gateways. This embodiment includes: reading data from a source network node to a memory space through a source network card; transferring information in the memory space It is transmitted to the target user mode protocol stack, and the target user mode protocol stack constructs the message of the target network protocol type according to the information of the memory space; the data in the memory is sent to the target network node through the target network card. When the network stacks are not equal and protocol conversion is required, there is no need to go through the kernel mode protocol stack, but the memory is the center. The user mode protocol stack shares the same memory data, and forwards the data in the memory in the user mode, reducing the memory data. CPU overhead caused by copying to the operating system protocol stack and switching from user mode to kernel mode. The implementation details of the data transmission method of this embodiment will be specifically described below. The following content is only provided for the convenience of understanding, and is not necessary for implementing this solution. A flowchart of the data transmission method according to the first embodiment of the present application is shown in FIG. 2 .

Step 201: Read data from the source network node to the memory space through the source network card.

In an example, before reading data from the source network node to the memory space through the source network card, a forwarding rule for data in the memory is configured according to the link between the source network node and the target network node, wherein the forwarding rule includes: source network node information and target network node information, that is, data matching the source network node information will be sent to the target network node.

In one example, the forwarding rule includes: memory space information, input information, and output information; the input information includes source network node information, and the output information includes: target network information; the source network node information includes: protocol, source information, and target network information Including protocol, action, target information.

Exemplarily, taking the mobile phone client remotely downloading data as an example, the server port 80 provides a web network access service externally, and requests data from the source network. When the requested data is transmitted to the data gateway configured with the data gateway service, the data gateway Enable the data gateway service according to the configuration. The data gateway is compatible with RDMA and TCP protocols, and supports forwarding memory data to two target systems and allocating memory buffers. For example, the memory space information in the configured forwarding rule is that the memory space address is 0x12323455, the memory space size is 16MB, the input information includes the source network node information, the source network node information includes the source network protocol RDMA, and the source information, namely nodes node1, node2 , the output information includes the protocol tcp protocol of the target network, the action "trans", that is to say, data conversion is performed, the target information: 1000, and the data is forwarded to the tcp link numbered 1000. The above forwarding rules are configured by the json file. The forwarding rules of the memory , the configured forwarding rules are as follows:

{

"Memory Space":{"Address":"0x12323455","Size":16MB},

"enter":

{"Protocol":"rdma","Source Information":["node1","node2"]},

"output":

[

{"protocol":"tcp","action":"trans","target information":1000},

]

}

The rule specifies the corresponding memory space and size. As shown in Figure 3, data is read from node1 and node2 into the memory space, and the output stream is 1: forward data to the tcp link numbered 1000. The link is established in advance. After the link is established, the forwarding rule is configured. When the data needs to be forwarded, it can be forwarded directly according to the forwarding rule, thereby improving the forwarding efficiency.

This embodiment creates a memory space according to the configured forwarding rules, obtains data from the source network node node1 or node2 through the source network card RNIC, and reads a memory space with an address of 0x12323455 and a size of 16MB.

Step 202: Determine the target transmission protocol of the next hop according to the preset logic, and directly transmit the information of the memory space to the next hop data transmission protocol stack, that is, obtain the target network matching the source network node according to the preset forwarding rule node information, and determine the target user mode protocol stack according to the matching target network node information, forward the memory space information to the determined target user mode protocol stack, and the target user mode protocol stack constructs the target network protocol type according to the memory space information. The message carries the data address in the memory space. In one example, the data information (including source address and protocol type) in the memory is passed to the data gateway service of the user-mode protocol stack, and the next-hop routing information (including destination address and protocol type) of the data is obtained for subsequent Forwards the data in memory to the destination address according to the new protocol.

Exemplarily, the target user-mode TCP/IP protocol stack in this embodiment refers to the realization of the RDMA software stack, and based on the DPDK (Data Plane Development Kit) framework, the user-mode TCP/IP protocol stack is customized and transformed. Define the send queue SQ (Send Queue) queue, CQ (Complet Queue) queue and related memory commands. The data gateway service executes the write_cmd command to send messages to SQ, carrying memory information, target information, etc. The software protocol stack, that is, the user-mode TCP/IP protocol stack reads the SQ queues and runs them one by one, and the write_cmd instruction triggers the TCP sending process. The customized user-mode TCP/IP protocol stack uses header and data separation to construct TCP packets. Each mbuf packet carries the header content and data pointer, and data points to the memory data to be sent. The network card detects the to-be-sent message, and sends the header and data into a complete message to the network. After the memory data is sent, the protocol stack stores the completion event CE (complet event) in the CQ queue, and the gateway service checks the CQ queue to obtain the sending command. As a result, DB (Door Bell) signaling is supported here.

Step 203: Send the data in the memory space to the target network node through the target network card based on the message of the target network protocol type.

In an example, after the data is sent to the target network node through the target network card, an end event is stored in the completion queue; the sending result is obtained according to the end event in the completion queue.

Following the above example, according to the carried target message, for example, the target information "1000" in the forwarding rule configured above, the data is sent to the ENIC network node through the target network card ENIC. That is to say, the network card detects the message to be sent, and sends the complete message composed of header and data to the network, and transmits it to the ENIC network node through the network. After the memory data is sent, the protocol stack stores the completion event CE (complete event) in the CQ queue, and the gateway service checks the CQ queue to obtain the result of sending the command. DB (Door Bell) signal notification is supported here.

In this embodiment, a completion queue storage end event is defined in the user protocol stack. Compared with the related art, which needs to wait for the transmission result to be returned after the network card transmission is completed, this embodiment can synchronously perform transmission and asynchronous acquisition of the transmission result. In addition, the defined user-mode TCP/IP protocol stack is implemented with reference to the RDMA software stack, so that the custom user-mode TCP/IP protocol stack and the RDMA protocol stack are more compatible.

The above all take the source network card as the RNIC network card, the source network node as the RDMA network node, the target network card as the ENIC network card, and the target network node as the Ethernet node as an example, that is, the RDMA network node transmits data through the RNIC network card, and after the data is processed by the data gateway, Transmission to the Ethernet node through the ENIC network card. That is to say, after reading the data from the RDMA network node to the memory space through the RNIC network card, the information of the memory space is transmitted to the user-mode TCP/IP protocol stack, and the user-mode TCP/IP protocol stack constructs the TCP message to carry the memory. The data address in the space is then transmitted to the Ethernet node through the ENIC network card.

The data transmission is bidirectional, and the configuration file can also be changed. When the source network card is an ENIC network card, the source network node is an Ethernet node, the target network card is an RNIC network card, and the target network node is an RDMA network node, the data obtained by the ENIC network card is passed through After processing by the data gateway, it is transmitted to the RDMA network node. That is to say, after reading the data from the Ethernet node to the memory space through the ENIC network card, the information of the memory space is passed to the user-mode RDMA protocol stack, and the user-mode RDMA protocol stack constructs RDMA packets to carry the data in the memory space. The address is then transmitted to the Ethernet node through the RNIC network card.

This embodiment implements multiple user-mode protocol stacks, ie, RDMA protocol stacks, and shares data in the memory with the user-mode TCP/IP protocol stacks. It solves the problem that the current RDMA protocol network element cannot be connected with the traditional Ethernet node.

In one example, after the processed data is sent to the target network node through the target network card, the completed data of data forwarding in the memory space is obtained. If the completed data of the data in the memory space is equal to the number of data output streams, the The memory space is recovered, and the number of output streams is determined according to the output information in the forwarding rule, and the output information includes the target network node information.

Continuing from the above example, the created memory space is regarded as a directed graph node, and the input and output streams of the data node are configured according to business characteristics, that is, input information and output information. Read the data and forward it to the client, so the data node is configured with an in-degree of 1 and an out-degree of 1, the input stream is the RDMA network, the output stream is the TCP network, and the data gateway records the number of forwarding completions of the node. When the number of node forwarding is equal to When the degree value is out, the memory space of the node can be recycled and reused. The memory space can be reclaimed in time.

In this embodiment, the source network card is used to read data from the source network node into the memory space; the information in the memory space is transmitted to the target user mode protocol stack, and the message of the target network protocol type constructed by the target user mode protocol stack is obtained. Carry the data address in the memory space, and send the data in the memory to the target network node through the target network card based on the message of the target network protocol type, so that there is no need to go through the kernel mode protocol stack. The memory data is forwarded in the user mode, reducing the CPU overhead caused by copying the memory data to the operating system protocol stack and switching the kernel mode from the user mode, and has better data forwarding performance.

The second embodiment of the present application relates to a data transmission method. The second embodiment of the present application is roughly the same as the first embodiment, and the main difference is that: the forwarding rule further includes local storage information, and according to the local storage information configured in the forwarding rule Information transfers the data to local storage. The data gateway can be compatible with the conversion of multiple protocols, and multiple user mode protocol stacks share the memory space, which can realize both local storage and data forwarding in the network. The flow chart of the second embodiment of the present application is shown in FIG. 4 . .

Step 401: Read data from the source network node to the memory space through the source network card.

Exemplarily, before reading data from the source network node to the memory space through the source network card, taking the mobile phone client remotely downloading data as an example, the server port 80 provides web access services externally, and requests data from the source network. When the data is transmitted to the data gateway configured with the data gateway service, the data gateway service is enabled according to the configuration. The data gateway is compatible with three protocols: RDMA, TCP, and nvme (Non-Volatile Memory Express, non-volatile memory standard). Data is forwarded to three target systems. The forwarding rules configured in the data gateway include memory space information, and the memory buffer is allocated according to the memory space information. For example, if the memory space information is the address "0x12323455" and the size is 16MB, then the starting address is "0x12323455" and the memory space is 16MB in size. Buffer, forwarding rules also include input information and output information, input information includes protocol RDMA, source information is node1, node2; output information includes local storage information, such as protocol nvme, action write, target information nvme, and output information also includes target information Network node information, for example, the protocol is tcp protocol, the action is trans, the target information is 1000, and the forwarding rules of the memory are configured by the json file, as follows:

{

"Memory Space":{"Address":"0x12323455","Size":16MB},

"enter":

{"Protocol":"rdma","Source Information":["node1","node2"]},

"output":

[

{"protocol":"tcp","action":"trans","target information":1000},

{"protocol":"nvme","action":"write","target information":"nvme"}

]

}

The rule specifies that the corresponding memory reads data from node1 and node2, and the output stream is divided into two channels: forwarding to the tcp link numbered 1000, writing to the nvme device, and the file system needs to provide address information. The output information includes two kinds of information, one is the output information for transmitting data to the network node, and the other is the output information for writing the data to the local storage, for example, the nvme device.

In step 402, the data in the memory is transferred to the NVMe protocol stack according to the local storage information configured in the forwarding rule, and then written to the local storage.

Exemplarily, this embodiment transmits data according to the local storage information configured in the forwarding rule, that is, "protocol": "nvme", "action": "write", and "target information": "nvme" configured in the output. to local storage nvme.

Step 403 , according to the forwarding rule, transfer the data in the memory to the user-mode TCP protocol stack, and then send the processed data to the target network node through the target network card.

In one example, step 402 and step 403 may be performed in parallel.

Step 403 is substantially the same as steps 202 to 203 in the first embodiment of the present application, and details are not repeated here.

In one example, after the processed data is sent to the target network node through the target network card, the completed data of data forwarding in the memory space is obtained. If the completed data of the data in the memory space is equal to the number of data output streams, the The memory space is reclaimed, and the number of output streams is determined according to the output information in the forwarding rule, and the output information includes the target network node information and the local storage information.

Continuing from the above example, the created memory space is regarded as a directed graph node, and the input and output streams of the data node are configured according to business characteristics, that is, input information and output information. In this embodiment, the master node needs to read data from other nodes. , forwarded to the client, so the data node is configured with an in-degree of 1 and an out-degree of 2, the input stream is the RDMA network, the output stream is the TCP network and nvme storage, and the service gateway records the number of forwarding completions of the node. When the number of node forwarding is equal to When the out-degree value is reached, that is, when the number of completed data forwarding in the memory space is equal to the number of data output streams, the memory space of the node can be recycled and reused. The memory space can be reclaimed in time.

The data source in this embodiment is any one of Ethernet network ENIC and RDMA network RNIC, and the data target is any one of Ethernet, RDMA network, and NVMe storage devices, including but not limited to other types of transmission protocols, such as : iSCSI, NVMe-oF, VirtIO, etc.

The schematic diagram of the data transmission network in this embodiment is shown in Figure 5. The data gateway synchronizes data from other nodes in the cluster to the MR (Memeory Regin) space of the node through the RNIC; the data gateway forwards the data to the user-mode TCP protocol stack and User-mode nvme protocol stack, write the data processed by the user-mode nvme protocol stack into the nvme storage device, and transmit the packets processed by the user-mode TCP/IP protocol stack to the nodes of the Ethernet network. Obtain data through the network card and process the obtained data.

The data gateway in this embodiment includes: a source network card, a target network card, and a user-mode protocol stack; the source network card is used to read data from the source network node to the memory space, and the target user-mode protocol stack is used to encapsulate the data in the memory into a target protocol message, the target network card is used to send the data encapsulated by the target user mode protocol stack to the target network node.

It should be noted that the traditional Ethernet transmission protocols TCP and UDP are defined for data packets and cannot be forwarded through the data gateway service. The data transmission method in this embodiment customizes the user-mode transmission protocol stack. The memory is the operation object, which can be seamlessly connected to the data gateway service, so that the memory data can be directly connected to the ENIC network card channel, and the network card hardware DMA method reads the data and completes the sending process.

The data transmission method in this embodiment is compatible with network protocols and local storage protocols, and can realize zero-copy forwarding of data between network nodes and local storage according to rules.

The steps of the above various methods are divided only for the purpose of describing clearly. During implementation, they can be combined into one step or some steps can be split and decomposed into multiple steps. As long as the same logical relationship is included, they are all within the protection scope of this patent. ;Adding insignificant modifications to the algorithm or process or introducing insignificant designs, but not changing the core design of the algorithm and process are all within the scope of protection of this patent.

The third embodiment of the present application relates to a data transmission device, as shown in FIG. 6 , including: a data reading module 601 for reading data from a source network node to a memory space through a source network card; a data forwarding module 602 for using For forwarding the information in the memory space to the target user mode protocol stack, the target user mode protocol stack constructs a message of the target network protocol type according to the information in the memory space, and the message carries the data address in the memory space; the data sending module 603, using It is used to send the data in the memory space to the target network node through the target network card based on the message of the target network protocol type.

In an example, the data reading module 601 is further configured to configure a forwarding rule for data in memory according to the link between the source network node and the target network node, wherein the forwarding rule includes: source network node information and target network node information Network node information. The data forwarding module 602 is further configured to obtain the target network node information matching the source network node according to the forwarding rule, and determine the target user mode protocol stack according to the matched target network node information, and forward the information of the memory space to the determined target user mode. protocol stack.

In an example, the forwarding rule further includes: local storage information, and the data sending module 603 is further configured to transmit the data in the memory space to the local storage according to the local storage information configured in the forwarding rule.

In one example, the data sending module is further configured to obtain the number of completed data forwarding in the memory space; if the number of completed data forwarding in the memory space is equal to the number of data output streams, the memory space For recycling, the number of the output streams is obtained according to the number of target network nodes and the number of local storage nodes.

In one example, the source network card is an RNIC network card, the source network node is an RDMA network node, the target network card is an ENIC network card, the target network node is an Ethernet node, and the data processing module 602 is further configured to The information is transmitted to the user-mode TCP/IP protocol stack, and the user-mode TCP/IP protocol stack constructs the TCP message to carry the data address in the memory space.

In one example, the source network card is an ENIC network card, the source network node is an Ethernet node, the target network card is an RNIC network card, the target network node is an RDMA network node, and the data processing module 602 is further configured to transmit the information of the memory space. To the user-mode RDMA protocol stack, the user-mode RDMA protocol stack constructs an RDMA message to carry the data address in the memory space.

In one example, the user-mode TCP/IP protocol stack includes: a completion queue; the data sending module 603 is further configured to store an end event in the completion queue after sending the data in the memory space to the target network node through the target network card; The sending result is acquired according to the end event in the completion queue.

It is not difficult to find that this embodiment is a system example corresponding to the first embodiment, and this embodiment can be implemented in cooperation with the first embodiment. The relevant technical details mentioned in the first embodiment are still valid in this embodiment, and are not repeated here in order to reduce repetition. Correspondingly, the related technical details mentioned in this embodiment can also be applied to the first embodiment.

It is worth mentioning that each module involved in this embodiment is a logical module. In practical applications, a logical unit may be a physical unit, a part of a physical unit, or multiple physical units. A composite implementation of the unit. In addition, in order to highlight the innovative part of the present application, this embodiment does not introduce units that are not closely related to solving the technical problem raised by the present application, but this does not mean that there are no other units in this embodiment.

The fourth embodiment of the present application relates to a network device, as shown in FIG. 7 , including: at least one processor 701; and,

memory 702 in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one The processor is capable of executing the above-described data transmission method.

The memory and the processor are connected by a bus, and the bus may include any number of interconnected buses and bridges, and the bus connects one or more processors and various circuits of the memory. The bus may also connect together various other circuits, such as peripherals, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides the interface between the bus and the transceiver. A transceiver may be a single element or multiple elements, such as multiple receivers and transmitters, providing a means for communicating with various other devices over a transmission medium. The data processed by the processor is transmitted on the wireless medium through the antenna, and further, the antenna also receives the data and transmits the data to the processor.

The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interface, voltage regulation, power management, and other control functions. Instead, memory may be used to store data used by the processor in performing operations.

The fifth embodiment of the present application relates to a computer-readable storage medium storing a computer program. The above method embodiments are implemented when the computer program is executed by the processor.

That is, those skilled in the art can understand that all or part of the steps in the method for implementing the above embodiments can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium and includes several instructions to make a device ( It may be a single chip microcomputer, a chip, etc.) or a processor (processor) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes .

Those of ordinary skill in the art can understand that the above-mentioned embodiments are specific examples for realizing the present application, and in practical applications, various changes in form and details can be made without departing from the spirit and the spirit of the present application. scope.

Claims

A method of data transmission, comprising:

Read data from the source network node to the memory space through the source network card;

The information of the memory space is forwarded to the target user mode protocol stack, and the target user mode protocol stack constructs a message of the target network protocol type according to the information of the memory space; the message carries the data in the memory space. data address;

The data in the memory space is sent to the target network node through the target network card based on the message of the target network protocol type.
The data transmission method according to claim 1, wherein before the data is read from the source network node to the memory space through the source network card, the method further comprises:

Configure a forwarding rule for data in memory according to the link between the source network node and the target network node, wherein the forwarding rule includes: source network node information and target network node information;

The forwarding of the information of the memory space to the target user mode protocol stack includes:

According to the forwarding rule, obtain target network node information that matches the source network node, and determine a target user mode protocol stack according to the matched target network node information;

The information of the memory space is forwarded to the determined target user mode protocol stack.
The data transmission method according to claim 2, wherein the forwarding rule further comprises: storing information locally;

After the data is read from the source network node to the memory space through the source network card, the method further includes:

The data in the memory space is transmitted to the local storage according to the local storage information configured in the forwarding rule.
The data transmission method according to claim 3, wherein after sending the processed data to the target network node through the target network card, the method further comprises:

Obtain the number of completed data forwarding in the memory space;

If the number of completed data forwarding in the memory space is equal to the number of data output streams, the memory space is reclaimed, wherein the number of the data output streams is determined according to the output information in the forwarding rule, and the The output information includes: the target network node information and the local storage information.
The data transmission method according to any one of claims 1 to 4, wherein the source network card is an RNIC network card, the source network node is an RDMA network node, the target network card is an ENIC network card, and the target network node is an Ethernet node;

Alternatively, the source network card is an ENIC network card, the source network node is an Ethernet node, the target network card is an RNIC network card, and the target network node is an RDMA network node.
The data transmission method according to claim 5, wherein the user mode protocol stack comprises: an RDMA protocol stack and a user mode TCP/IP protocol stack;

If the source network card is an RNIC network card, the source network node is an RDMA network node, the target network card is an ENIC network card, and the target network node is an Ethernet node, then the information of the memory space is forwarded to the target network User mode protocol stack, the target user mode protocol stack constructs the message of the target network protocol type according to the information of the memory space, including:

The information of the memory space is delivered to the user state TCP/IP protocol stack, and the user state TCP/IP protocol stack constructs a TCP message to carry the data address in the memory space;

If the source network card is an ENIC network card, the source network node is an Ethernet node, the target network card is an RNIC network card, and the target network node is an RDMA network node, the forwarding of the memory space information to the target User mode protocol stack, the target user mode protocol stack constructs the message of the target network protocol type according to the information of the memory space, including:

The information of the memory space is transferred to the user-mode RDMA protocol stack, and an RDMA message is constructed by the user-mode RDMA protocol stack to carry the data address in the memory space.
The data transmission method according to claim 6, wherein the user-mode TCP/IP protocol stack comprises: a completion queue;

After sending the data in the memory space to the target network node through the target network card based on the message of the target network protocol type, the method further includes:

Store the end event into the completion queue;

The sending result is acquired according to the end event in the completion queue.
A data transmission device, comprising:

The data reading module is used to read data from the source network node to the memory space through the source network card;

A data forwarding module, configured to forward the information of the memory space to the target user mode protocol stack, and the target user mode protocol stack constructs a message of the target network protocol type according to the information of the memory space, and the message carries a data address in the memory space;

The data sending module is configured to send the data in the memory space to the target network node through the target network card based on the message of the target network protocol type.
A network device comprising:

at least one processor; and,

a memory communicatively coupled to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform any one of claims 1 to 7 data transfer method.
A computer-readable storage medium storing a computer program, when the computer program is executed by a processor, the data transmission method according to any one of claims 1 to 7 is implemented.