WO2023005723A1 - Packet transmission method and communication apparatus - Google Patents

Abstract

Disclosed in embodiments of the present application are a packet transmission method and a communication apparatus, for use in improving a packet transmission rate. The method in the embodiments of the present application comprises: a gateway parses a first packet of a first network device to obtain an IP address of a second network device; matches in a lookup table an LID corresponding to the IP address of the second network device; encapsulates a second packet header comprising the LID of the second network device to the first packet from which a first packet header is stripped, so as to form a second packet, the LID being stored in a local routing header of the second packet; and then sends the second packet according to the LID of the second network device.

Description

A message transmission method and communication device

This application claims the priority of the Chinese patent application filed with the State Intellectual Property Office of China on July 30, 2021, with the application number 202110872533.9, and the title of the invention is "A Message Transmission Method and Communication Device", the entire contents of which are incorporated by reference in this application.

technical field

The embodiments of the present application relate to the communication field, and in particular, to a message transmission method and a communication device.

Background technique

With the rapid increase of data scale, the computing power demand of application performance on processing system has expanded exponentially, and the demand for high performance computing (high performance computing, HPC) cluster applications has increased dramatically. In the HPC interconnection network, Ethernet and infiniband technology (infiniband, IB) network clusters account for 78%, far exceeding other interconnection networks. Remote direct memory access (RDMA) technology first appeared on the IB network. IB has been the first choice for supercomputing interconnection since its inception due to its high performance and low latency, but the remote direct access (RDMA over Converged Ethernet) based on Ethernet Version2, RoCEv2) network has been adopted by more and more supercomputing interconnections because it is fully compatible with the Ethernet protocol (internet protocol, IP) network and supports the RDMA protocol.

The Ethernet IP message is transmitted on the IB network through the gateway device using the Internet protocol over infiniband (IPoIB) technology running on the IB.

However, because the Ethernet IP message is directly encapsulated in the IB message through the tunneling technology, the message still passes through the kernel copy and software protocol stack, which cannot realize the remote direct access effect of the IB network and affects the transmission efficiency.

Contents of the invention

Embodiments of the present application provide a message transmission method and a communication device, which are used to increase a message transmission rate.

The first aspect of the embodiment of the present application provides a message transmission method, the method includes: the gateway receives the first message from the first network device, the first message header of the first message includes the Internet address of the second network device Protocol IP address, the second network device is the destination network device for the first network device to transmit the message; the gateway determines the local identifier LID of the second network device according to the IP address of the second network device in combination with the lookup table, and the lookup table includes the IP address The association relationship with the LID; the gateway strips the first packet header of the first packet and encapsulates the second packet header to obtain the second packet. The second packet header includes the local routing header, and the local routing header includes the first packet header. The LID of the second network device; the gateway sends the second message according to the LID of the second network device.

In the first aspect above, the gateway parses the first message of the first network device to obtain the IP address of the second network device, matches the LID corresponding to the IP address of the second network device in the lookup table, and includes the LID of the second network device The second packet header is encapsulated into the first packet after the first packet header is stripped to generate a second packet, and then the second packet is sent according to the LID of the second network device. The second packet is an ordinary IB format message, the LID of the second network device is stored in the local routing header of the second message, the transmission of the second message on the IB network does not need to be copied by the kernel, and the effect of remote direct access to the IB network can be realized, which improves the reporting efficiency. file transfer efficiency.

In a possible implementation manner, the above steps where the gateway receives the first message from the first network device include: the gateway receives the first message in a large-capacity buffer according to the credit flow control mechanism; the gateway sends the first message to the first by suspending the message The network device feeds back the status information of the large-capacity buffer, so that the first network device adjusts packet transmission.

In the above possible implementation manner, the gateway can use the credit flow control mechanism of the IB network and combine a large-capacity buffer to receive messages from the first network device, and the large-capacity buffer can receive more flight messages , the gateway can also indicate the status information of the buffer to the first network device by suspending the message, so that the first network device can adjust the transmission of subsequent messages, and the first network device can reduce or stop sending messages accordingly, which can avoid Transmission congestion.

In a possible implementation manner, before the gateway receives the first message from the first network device in the above step, the method further includes: the gateway applies for a LID from the subnet manager according to the route change, and the route change instructs the first network device to join the network The gateway receives the LID of the first network device; the gateway obtains a response message from the second network device, and the response message includes the LID and IP address of the second network device; the gateway according to the IP address and LID of the first network device, and the second 2. The IP address and LID of the network device update the lookup table.

In the above possible implementation manner, the gateway can apply for a LID for the subnet manager on the IB side of the first network device accessing the network, and obtain the response message of the second network device according to the broadcast message of the first network device, according to In response to the IP address and LID of the second network device in the message, a lookup table is generated or updated to improve the feasibility of the solution.

In a possible implementation manner, the gateway in the above steps applying for a LID from the subnet manager according to the route change includes: the gateway receives an address resolution protocol ARP message from the first network device, and the ARP message includes the IP address of the first network device and The IP address of the second network device; the gateway applies to the subnet manager for the LID of the first network device according to the ARP message.

In the above-mentioned possible implementation mode, the broadcast message can be an ARP message, and the route change can also be determined by the ARP message, and the gateway converts the ARP message into an ARP message in IB format, and sends it to the second network device, The second network device feeds back an ARP response message, and the ARP response message includes the IP address and LID of the second network device, which improves the feasibility of the solution.

In a possible implementation manner, in the above steps, the gateway strips the first packet header of the first packet and encapsulates the second packet header, so that after obtaining the second packet, the method further includes: the gateway updates the second packet Invariant cyclic redundancy check code (invariant cyclic redundancy check, ICRC) and variable cyclic redundancy check code (variant cyclic redundancy check, VCRC).

In the above possible implementation manners, after the gateway converts the first message into the second message, the gateway needs to modify the ICRC and VCRC of the second message to increase the error checking capability.

In a possible implementation manner, the first message is an Ethernet message, and the second message is an IB message.

In a possible implementation manner, the Ethernet packet includes an Ethernet header, an IP header, a UDP header, an IB transmission header, an IB payload, an ICRC, and a cyclic redundancy check code (cyclic redundancy check, CRC).

In a possible implementation manner, the IB packet includes a local routing header, an IB transport header, an IB payload, an ICRC, and a VCRC.

The second aspect of the embodiment of the present application provides a message transmission method, including: the gateway receives the third message from the first network device, the third message header of the third message includes a local route header, and the local route The header includes the local identifier LID of the second network device, and the second network device is the destination network device for the first network device to transmit the message; the gateway determines the Internet protocol IP of the second network device according to the LID of the second network device and a lookup table address, the lookup table includes the association relationship between the IP address and the LID; the gateway strips the third packet header of the third packet, and encapsulates the fourth packet header to obtain the fourth packet, and the fourth packet header includes the second The IP address of the network device; the gateway sends the fourth message according to the IP address of the second network device.

In the second aspect above, the gateway parses the first packet of the first network device to obtain the LID of the second network device, matches the IP address corresponding to the LID in the lookup table, and sends the second packet including the IP address of the second network device The header is encapsulated, and the second packet is generated from the first packet after stripping the first packet header, and then the second packet is sent according to the LID of the second network device. The transmission of the second packet on the Ethernet network does not require After the kernel is copied, the effect of remote direct access to the Ethernet network can be realized, and the efficiency of message transmission can be improved.

In a possible implementation manner, the method also includes: the gateway obtains the queue pair number (queue pair number, QPN) of the first network device and the QPN of the second network device according to the link establishment message; The QPN of the second network device and the QPN of the second network device obtain the user datagram protocol UDP port number of the second network device, the lookup table also includes the association relationship between the QPN, the UDP port number, the IP address and the LID, and the fourth packet header also includes the second The MAC address and UDP port number of the media access control layer of the network device, and the MAC address of the second network device is obtained by broadcasting the IP address of the second network device.

In the above possible implementation manner, when the first network device and the second network device establish a transmission link, the gateway can also record the QPN of the first network device and the second network device, and pass the first network device and the second network device The QPN of the device calculates the UDP port number of the second network device, and the gateway can also send a broadcast message to the RoCE network according to the IP address of the second network device, so that the second network device can feedback the MAC address and send the second network device’s The MAC address and UDP port number are encapsulated into the second message, so that the gateway can transmit the second message according to the IP address, MAC address and UDP port number of the second network device, thereby improving the reliability of message transmission.

In a possible implementation manner, the above steps where the gateway sends the fourth message according to the IP address of the second network device include: the gateway sends the fourth message in a large-capacity buffer according to the credit flow control mechanism and the IP address of the second network device. The gateway feeds back the status information of the large-capacity buffer to the second network device by suspending the message, so that the second network device adjusts message transmission.

In a possible implementation manner, before the gateway receives the third message from the first network device in the above step, the method further includes: the gateway applies for a LID from the subnet manager according to the route change, and the route change instructs the second network device to join the network The gateway receives the LID of the second network device; the gateway obtains a response message from the first network device, and the response message includes the LID and IP address of the first network device; the gateway according to the IP address and LID of the first network device, and the second 2. The IP address and LID of the network device update the lookup table.

In a possible implementation manner, the gateway in the above steps applying for a LID from the subnet manager according to the route change includes: the gateway receives an address resolution protocol ARP message from the second network device, and the ARP message includes the IP address of the first network device and The IP address of the second network device; the gateway applies to the subnet manager for the LID of the second network device according to the ARP message.

In a possible implementation manner, in the above steps, the gateway strips the first packet header of the first packet and encapsulates the second packet header, so that after obtaining the second packet, the method further includes: the gateway updates the second packet The ICRC and CRC of the text.

In a possible implementation manner, the first message is an Ethernet message, and the second message is an IB message.

In a possible implementation manner, the Ethernet packet includes an Ethernet header, an IP header, a UDP header, an IB transmission header, an IB payload, an ICRC, and a CRC.

In a possible implementation manner, the IB message includes a local routing header, an IB transmission header, an IB payload, an ICRC, and a VCRC.

The third aspect of the embodiment of the present application provides a communication device, including: a receiving unit, configured to receive a first message from a first network device, the first message header of the first message includes the Internet address of the second network device Protocol IP address, the second network device is the destination network device for the first network device to transmit the message; the determination unit is used to determine the local identifier LID of the second network device according to the IP address of the second network device in combination with a lookup table, and search The table includes the association relationship between the IP address and the LID; the encapsulation unit is used to strip the first packet header of the first packet and encapsulate the second packet header to obtain the second packet, the second packet header includes the local A routing header, where the local routing header includes the LID of the second network device; a sending unit configured to send the second packet according to the LID of the second network device.

The communication device is configured to execute the method of the foregoing first aspect or any implementation manner of the first aspect.

The fourth aspect of the embodiment of the present application provides a communication device, including: a receiving unit, configured to receive a third message from the first network device, the third message header of the third message includes a local routing header, the The local routing header includes the local identifier LID of the second network device, and the second network device is the destination network device of the message transmitted by the first network device; the determination unit is used to determine the second network device according to the LID of the second network device in combination with a lookup table The Internet protocol IP address of the network device, the lookup table includes the relationship between the IP address and the LID; the encapsulation unit is used to strip the third message header of the third message and encapsulate the fourth message header to obtain the fourth message The fourth message header includes the IP address of the second network device; the sending unit is configured to send the fourth message according to the IP address of the second network device.

The communication device is configured to execute the method of the foregoing second aspect or any implementation manner of the second aspect.

The fifth aspect of the embodiment of the present application provides a communication device, including: a processor, a memory, and a communication interface. The processor is used to execute instructions stored in the memory, so that the communication device performs the first aspect or any of the first aspects. An optional method provided by the communication interface for receiving or sending indications. For specific details of the communication device provided in the fifth aspect, refer to the first aspect or any optional manner of the first aspect, and details are not repeated here.

The sixth aspect of the embodiment of the present application provides a communication device, including: a processor, a memory, and a communication interface, the processor is used to execute instructions stored in the memory, so that the communication device performs the second aspect or any of the second aspects. An optional method provided by the communication interface for receiving or sending indications. For specific details of the communication device provided in the sixth aspect, reference may be made to the second aspect or any optional manner of the second aspect, and details are not repeated here.

The seventh aspect of the embodiment of the present application provides a computer-readable storage medium, where a program is stored in the computer-readable storage medium, and when the computer executes the program, the first aspect or any optional method of the first aspect can be executed provided method.

The eighth aspect of the embodiment of the present application provides a computer-readable storage medium, the computer-readable storage medium stores a program, and when the computer executes the program, it executes the second aspect or any optional method of the second aspect provided method.

A ninth aspect of the embodiments of the present application provides a computer program product. When the computer program product is executed on a computer, the computer executes the method provided in the foregoing first aspect or any optional manner of the first aspect.

The tenth aspect of the embodiments of the present application provides a computer program product. When the computer program product is executed on a computer, the computer executes the method provided in the foregoing second aspect or any optional manner of the second aspect.

Description of drawings

Fig. 1 is the HPC system block diagram that the embodiment of the present application provides;

FIG. 2 is a schematic diagram of a short-distance scene inside a data center provided by an embodiment of the present application;

FIG. 3 is a schematic diagram of a long-distance scenario of a supercomputing intermediate interconnection provided by an embodiment of the present application;

FIG. 4 is a schematic diagram of an embodiment of a message transmission method provided by an embodiment of the present application;

FIG. 5 is a schematic diagram of another embodiment of the message transmission method provided by the embodiment of the present application;

FIG. 6 is a schematic structural diagram of a gateway provided by an embodiment of the present application;

FIG. 7 is a schematic diagram of another structure of the gateway provided by the embodiment of the present application;

FIG. 8 is a schematic diagram of another structure of the gateway provided by the embodiment of the present application;

FIG. 9 is a schematic structural diagram of a communication device provided by an embodiment of the present application;

FIG. 10 is another schematic structural diagram of a communication device provided by an embodiment of the present application;

FIG. 11 is a schematic structural diagram of a communication device provided by an embodiment of the present application;

FIG. 12 is a schematic diagram of another structure of a communication device provided by an embodiment of the present application.

Detailed ways

Embodiments of the present application provide a message transmission method and a communication device, which are used to increase a message transmission rate.

Embodiments of the present application are described below in conjunction with the accompanying drawings. Apparently, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Those of ordinary skill in the art know that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

The terms "first", "second" and the like in the specification and claims of the present application and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific sequence or sequence. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" and "having", as well as any variations thereof, are intended to cover a non-exclusive inclusion, for example, a process, method, system, product or device comprising a sequence of steps or elements is not necessarily limited to the expressly listed instead, may include other steps or elements not explicitly listed or inherent to the process, method, product or apparatus.

The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as superior or better than other embodiments.

The tables in this application can be split and merged, and are not limited thereto, and an example is only given here.

In addition, in order to better illustrate the present application, numerous specific details are given in the following specific implementation manners. It will be understood by those skilled in the art that the present application may be practiced without certain of the specific details. In some instances, methods, means, components and circuits well known to those skilled in the art have not been described in detail in order to highlight the gist of the present application.

With the rapid increase of data scale, the computing power demand of application performance on processing system has expanded exponentially, and the demand for high-performance computing (High Performance Computing, HPC) applications has increased dramatically. The main industries and applications of high-performance computing (HPC) are in scientific research institutions such as schools and research institutes. Petroleum sector, medical biology, computational chemistry and automotive and aerospace design, architectural structure design, 3D graphics computing and other fields. HPC refers to computing systems and environments that use many processors (part of a single machine) or several computers in a cluster (operating as a single computing resource). In the face of large-scale computing tasks, parallel algorithms are used to split and distribute a large task to different nodes in the cluster for parallel computing, and then aggregate the calculation results to quickly obtain the final result. It can be seen that the performance of the HPC system is not only related to the calculation of the computing nodes. Power and storage performance are also closely related to node interconnection network performance. Infiniband technology (infiniband, IB) has been the first choice for supercomputing interconnection since its birth due to its high performance and low latency. The Ethernet IP network supports the remote direct memory access (RDMA) protocol and is adopted by more and more supercomputing interconnections.

Wherein, the Ethernet message is a message transmitted in the Ethernet network, and is transmitted in the Ethernet network based on the IP address in the IP header.

The IB message is a message transmitted in the IB network. The IB network does not perceive the IP address. The network device in the IB network is assigned a local identifier (local identification, LID). The IB message is based on the LID in the local routing header in the IB network. in transmission.

At present, data centers have IB networks and RoCE networks coexisting and require interaction scenarios. For example, storage and computing use different high-speed interconnection networks, and equipment and devices that support the IB network and RoCE network interconnection are required. Figure 1 is a block diagram of the HPC system provided by the embodiment of this application. In the HPC cluster environment, high-performance RDMA network interconnection is used between computing nodes and storage nodes. Currently, the mainstream uses IB and RoCEv2 network interconnection, as shown in Figure 1. There is a mixed interconnection between IB and RoCE, and there is a situation where Ethernet packets are converted to IB packets.

Most of the HPC clusters use the IB and RoCE networks for networking, and there is a need for data transmission across the IB and RoCE networks. It is necessary to achieve efficient interworking between the IB network and the RoCE network.

The implementation of the IB and Ethernet RoCE conversion method in the embodiment of the present application can be used in efficient data interaction scenarios across the IB network and the RoCE network, mainly in short-distance scenarios within the data center and long-distance scenarios between supercomputing centers, as shown in Figure 2. The schematic diagram of the short-distance scene inside the data center provided in the embodiment of the application, the IB network and the RoCE network are transmitted through a gateway (gateway), and the IB message of the IB network is converted into an Ethernet message through the gateway for transmission on the RoCE network, or the RoCE network Ethernet packets are converted into IB packets through the gateway for transmission on the IB network. As shown in Figure 3, it is a schematic diagram of the long-distance scene of supercomputing intermediate interconnection provided by the embodiment of the present application, IB network 1, gateway 1, IB network 2 and gateway 2, and IB network 1 converts IB messages into Ethernet packets through gateway 1 The message is transmitted to the gateway 2, and the gateway 2 converts the Ethernet message into an IB message for transmission in the IB network 2. The embodiment of the present application takes a short-distance scene as an example.

In the prior art, the Ethernet IP message is transmitted on the IB network through a gateway device (GateWay) using the Internet protocol (internet protocol over infiniband, IPoIB) technology running on the IB, wherein the TCP/IP message The text is directly encapsulated in the IB message transmission through the tunnel technology, and the message still passes through the kernel copy and software protocol stack, which cannot take advantage of the IB network kernel bypass (kernel bypass) and zero copy (zero copy), and the delay is large and the CPU usage is high , Traditional packets are encapsulated by the IB network, which cannot take advantage of the high efficiency of the IB network. Especially for small message packets, the packet encapsulation efficiency is significantly reduced.

In order to solve the above problem, the embodiment of the present application provides a message transmission method, the method is as follows.

Please refer to FIG. 4. As shown in FIG. 4, it is an embodiment of a message transmission method provided by the embodiment of the present application. The method includes:

401. The first network device sends the first packet to the gateway.

In this embodiment of the present application, the first network device is a network device on the RoCE network side, and the second network device is a network device on the IB network side. Wherein, the network device on the RoCE network side needs to send data to the network device on the IB network side. Specifically, the first network device can send the first message including the data to the second network device through the gateway. The first message for Ethernet packets. Wherein, the first message includes a media access control (media access control, MAC) address of the first network device, an Internet protocol (internet protocol, IP) address, a user datagram protocol (user datagram protocol, UDP) port number, and , the MAC address and IP address of the second network device.

The message format of the first message is shown in Table 1 below, and the first message includes an Ethernet (ethernet, ETH) header (header), an IP header, a UDP header, an IB transmission (transport) header, an IB payload (payload) ), an invariant cyclic redundancy check code (invariant cyclic redundancy check, ICRC) and a cyclic redundancy check code (cyclic redundancy check, CRC) field, wherein, the ETH header stores the information of the first network device and the second network device MAC address, the IP header stores the IP addresses of the first network device and the second network device, the UDP header stores the UDP port number of the first network device, and the IB transport header field stores the QPN of the first network device and the second network device The QPN, ICRC field and CRC field are used to check the data in the frame to ensure the correctness of data transmission.

Table 1

402. The gateway determines the LID of the second network device according to the IP address of the second network device in combination with a lookup table.

In the embodiment of the present application, after the gateway receives the first message, it can analyze the first message to obtain the IP address of the second network device in the IP header, and then according to the IP address and LID in the lookup table Association relationship, matching the IP address of the second network device with the lookup table to obtain the LID of the second network device.

Optionally, before step 401, the gateway applies for a LID from the subnet manager according to the routing change; the gateway receives the LID of the first network device; the gateway obtains a response message from the second network device; The address and LID, and the IP address and LID of the second network device update the lookup table. Specifically, when the first network device is connected to the network, the gateway needs to add a routing path, that is, the gateway can determine the routing change, and the gateway can apply for a LID from the subnet manager, which is located in the IB network, and the subnet manager A LID may be allocated randomly, and the gateway may correspond the LID to the IP address of the first network device. The gateway can also receive a broadcast message from the first network, convert it into an IB message and forward it to the second network device, so that the second network device feeds back a response message, which includes the IP address of the second network device address and LID, the gateway can associate the IP address and LID of the first network device with the IP address and LID of the second network device, and store them in the lookup table, or update the lookup table.

Optionally, the gateway may apply for a LID from the subnet manager according to the route change, and the gateway receives an address resolution protocol (address resolution protocol, ARP) message from the first network device, and applies for the first subnet manager according to the ARP message. A LID of a network device. Specifically, the gateway can also recognize that there is a newly added terminal on the Ethernet side by receiving an ARP message from the Ethernet side or a message from other protocols. When receiving the ARP message, the gateway can obtain the IP address in the ARP message , and apply for a LID from the subnet manager for the first network device. Correspondingly, the gateway converts the ARP message into an ARP message on the IB side, and sends it to the second network device, and the response message sent by the second network device is an ARP response message.

Optionally, after the above-mentioned lookup table is generated or updated, aging settings can also be performed, and the source IP address and destination IP address in the lookup table can be used as the basis for judgment. source IP address and destination IP address, delete the association relationship between the source IP address and destination IP address in the lookup table, or if the source IP address and destination IP address are not received within a set time message, the lookup table enters the aging counting process, and when the aging counter reaches a preset value, the association between the source IP address and the destination IP address in the lookup table is deleted. In this embodiment, the deletion of the source IP address and the destination IP address The way of looking up the association relationship in the table is not limited. Specifically, if an Ethernet message containing a specific IP is received during the counting period of the aging counting process, the aging counting process is re-entered.

403. The gateway strips the first packet header of the first packet, and encapsulates the second packet header, to obtain the second packet.

In the embodiment of the present application, after the gateway obtains the LID of the second network device from the lookup table according to the IP address of the second network device, it can strip the first message header of the first message, wherein the first message includes the following For the ETH header, IP header, and UDP header in Table 1, the gateway can encapsulate the first packet after stripping the first packet header into the second packet header to form the second packet. The second packet header includes The LID of the second network device, the second message header is an IB message, and the conversion between the IB message and the Ethernet message is realized by using a hardware lookup table. The format of the second message is as shown in Table 2, and the Local Route Header (Local Route Header) is included in the second message header, and the Local Route Header field includes the LIDs of the first network device and the second network device.

Table 2

After the gateway strips the first message header and encapsulates the second message header, it can also update the UDP port number in the first message header and the QPN in the IB transport header to the lookup table.

The format of the lookup table can be shown in Table 3 below, Src represents the source device, Dst represents the destination device, and Src UDP port1 represents the UDP port number of the second network device.

table 3

Src IP1	Dst IP1	Src LID1	Dst LID1	Dst QPN1	Src QPN1	Src UDP port1
Src IP2	Dst IP2	Src LID2	Dst LID2	Dst QPN2	Src QPN2	Src UDP port2
Src IP3	Dst IP3	Src LID3	Dst LID3	Dst QPN3	Src QPN3	Src UDP port3
…	…	…	…	…	…	…
Src IPn	Dst IPn	Src LIDn	Dst LIDn	Dst QPNn	Src QPNn	Src UDP portn

Optionally, before the gateway sends the second message to the second network device, the gateway may also update the ICRC and variable cyclic redundancy check code (variant cyclic redundancy check, VCRC) of the second message. Specifically, after the gateway converts the first message into the second message, the gateway needs to modify the ICRC and VCRC of the second message to increase the code distance and error detection and correction capabilities of the entire coding system.

Realize the conversion of Ethernet messages into IB messages and update the ICRC and VCRC of the messages while recording the UDP port number, and also map the RoCEv2 type of service (TOS)/differentiated services code point (DSCP) The service level (service level) SL field of the IB message realizes the quality of service (quality of service, QOS) information transmission. DSCP uses the used 6 bits and the unused 2 bits in the service class TOS identification byte of each data packet IP header to distinguish the priority by encoding the value. Select 8 of the 16 virtual ports on the IB side to map with the 8 sending queues on the RoCE network. The value range of the DSCP field on the Ethernet side and the mapping method of the SL field on the IB side are shown in Table 4.

Table 4

RoCEv2 side DSCP (6bits)	IB side SL(4bits)	Priority
0-7	0	Priority 0
8-15	1	Priority 1
16-23	2	Priority 2
24-31	3	Priority 3
32-39	4	Priority 4
40-47	5	Priority 5
48-55	6	Priority 6
56-63	7	Priority 7

404. The gateway sends the second packet according to the LID of the second network device.

In the embodiment of the present application, after the gateway generates the second message, it can send the second message to the second network device in the IB network according to the LID of the second network device.

For the supercomputing center interconnection long-distance interconnection scenario, the local gateway and the remote gateway need to negotiate a common subnet manager, that is, the LID allocated by the subnet manager needs to be uniquely identified in the IB network where the local gateway and the remote gateway are located. The local gateway converts IB packets into Ethernet packets, and the IP address and MAC address can be configured on the local gateway.

The local gateway adopts a two-level flow control method for receiving packets from the remote gateway:

Optionally, the gateway receiving the first message from the first network device may be that the gateway receives the first message in a large-capacity buffer according to a credit (credit) flow control mechanism; the gateway sends the first message to the first The network device feeds back the status information of the large-capacity buffer, so that the first network device adjusts packet transmission. Specifically, the gateway adopts the original credit flow control mechanism of the IB network, and uses first-in-first-out queue (first input first output, FIFO) large-capacity buffer docking based on virtual lane (virtual lane, VL) granularity, and adopts configurable FIFO water Line, while monitoring the internal storage of FIFO in real time, the large-capacity buffer can receive more flight messages. When the remote Ethernet side transmits a large amount of data to the local gateway, the local gateway transmits the FIFO status information to the Ethernet port of the remote gateway through the Ethernet flow control pause message, and at the same time analyzes the Pause message combined with the waterline setting to determine the congestion situation of the peer end. The originator makes adjustments to ensure efficient transmission and no congestion at the peer end. Concretely, can use the priority-based flow control (priority-based flow control, PFC) function of port, can carry out flow control to message based on 802.1P priority, change the pause time in PFC to buffer (buffer) stacking situation.

In the technical solution of the embodiment of the present application, the gateway parses the first message of the first network device to obtain the IP address of the second network device, matches the LID corresponding to the IP address of the second network device in the lookup table, and includes the second network device The second packet header of the LID is encapsulated into the first packet after stripping the first packet header to generate a second packet, and then the second packet is sent according to the LID of the second network device, and the second packet is in The transmission of the IB network does not need to be copied by the kernel, which can realize the effect of remote direct access to the IB network and improve the efficiency of message transmission.

Described above is the method that the gateway converts the Ethernet message into an IB message, and the following describes the method that the gateway converts the IB message into an Ethernet message. The first network device is a network device of the IB network, and the second network device is an Ethernet network. Internet equipment.

Please refer to FIG. 5 . FIG. 5 is a schematic diagram of another embodiment of a message transmission method provided by the embodiment of the present application, and the method is as follows.

501. The first network device sends the third packet to the gateway.

In the embodiment of the present application, the third message is an IB message, and the format of the IB message can refer to the format of the IB message in step 403. The third message header of the third message includes a Local Route Header, and the Local Route The Header includes the LID of the first network device and the LID of the second network device, and the QPN of the first network device and the QPN of the second network device.

502. The gateway determines the Internet Protocol IP address of the second network device according to the LID of the second network device in combination with a lookup table.

In step 502, the manner of searching for the IP address according to the LID and the manner of updating the lookup table can refer to the relevant description of the manner of searching for the LID according to the IP address and the manner of updating the lookup table in step 402, which will not be repeated here.

503. The gateway strips the third packet header of the third packet, and encapsulates the fourth packet header, to obtain the fourth packet.

In the embodiment of the present application, the fourth message header includes ETH header, IP header, UDP header, IB transport header, IB payload and CRC fields, the gateway receives the IB message, and parses the source LID and the source LID of the LRH field inside the IB message The destination LID and destination QPN, through the source LID and destination LID of the message, and the destination QPN matching lookup table, realize the source IP address, destination IP address and UDP port number of the Ethernet header IP header of the Ethernet message, and at the same time set the IB transport header and the IB payload are directly encapsulated into the corresponding location field of RoCEv2, wherein the gateway can also send a broadcast message to the RoCE network according to the destination IP address, so that the second network device can feedback the MAC address, then the MAC address is the destination MAC address, and the The destination MAC address is stored in the ETH header field. Based on the hop-by-hop nature of the MAC address in the ETH header of the Ethernet message, the MAC address of the Ethernet side interface of the gateway can be used as the source MAC address, and the ICRC and CRC fields of the message are updated at the same time and encapsulated into the corresponding check field. At this point, the gateway implements IB IB packets on the Ethernet side and Ethernet packets on the Ethernet side are converted to each other. For the fourth message, refer to the format of the first message in FIG. 4 , which will not be repeated here.

Optionally, the gateway obtains the QPN of the first network device and the QPN of the second network device according to the link establishment message; the gateway obtains the UDP port number of the second network device according to the QPN of the first network device and the QPN of the second network device . Specifically, when the first network device establishes a connection with the second network device, the first network device may send a link establishment message to the gateway, and the gateway converts the link establishment message into an IB format and sends it to the second network device, and the second The network device then feeds back a link establishment response message; it may also be that the second network device sends a link establishment message to the gateway, and the gateway converts the link establishment message into an Ethernet format and sends it to the first network device, and the first network device feeds back the establishment link message. Chain response message. The gateway can obtain the QPN of the first network device and the QPN of the second network device according to the link establishment message, and calculate the second link based on the QPN of the first network device and the QPN of the second network device according to the mapping relationship between the QPN and the UDP port number. The UDP port number of the second network device, the gateway can store the corresponding relationship between the QPN of the first network device, the QPN of the second network device, and the UDP port number of the second network device in the lookup table, when the gateway is the second network device After the LID is allocated, the lookup table may also include the association relationship among IP address, LID, QPN and UDP port number. Then the gateway can determine the MAC address of the first network device and the MAC address of the second network device, the IP address of the first network device and the IP address of the second network device required by the fourth message header according to the third message, The QPN of the first network device, the QPN of the second network device, and the UDP port number of the second network device.

504. The gateway sends the fourth packet according to the IP address of the second network device.

In the embodiment of the present application, the gateway may transmit the fourth message according to the IP address of the second network device in the fourth message according to the transmission mode of the Ethernet message.

Optionally, the gateway sends the fourth message according to the IP address of the second network device. Two-level flow control transmission can be used. The gateway sends the fourth message in a large-capacity buffer according to the credit flow control mechanism and the IP address of the second network device. The gateway feeds back the status information of the large-capacity buffer to the second network device by suspending the message, so that the second network device adjusts message transmission. For the relevant description of the two-stage flow control transmission, reference may be made to the relevant description in step 304, which will not be repeated here.

In the technical solution of the embodiment of the present application, the gateway parses the first message of the first network device to obtain the LID of the second network device, matches the IP address corresponding to the LID in the lookup table, and includes the first packet of the IP address of the second network device The second message header is encapsulated, and the second message is generated in the first message after stripping the first message header, and then the second message is sent according to the LID of the second network device, and the transmission of the second message on the Ethernet network It does not need to be copied by the kernel, and can realize the remote direct access effect of the Ethernet network, which improves the efficiency of message transmission.

The structure of the gateway in the embodiment of the present application can be a schematic diagram of a gateway structure as shown in Figure 6. The gateway includes a switching chip and a processing chip, wherein the switching chip mainly implements basic forwarding functions, and the processing chip can be a CPU, FPGA, etc. Responsible for the establishment and maintenance of lookup tables required for the conversion of different protocol messages and the conversion of different protocol messages.

The structure of the gateway in the embodiment of the present application may be another schematic structural diagram of the gateway as shown in FIG. For the basic forwarding function, the processing module is responsible for the establishment and maintenance of the lookup tables required for the conversion of different protocol messages and the conversion of different protocol messages.

In the embodiment of the present application, the structure of the processing chip can refer to the schematic structural diagram of the processing chip shown in FIG. and management mods.

Among them, the Ethernet interface is used to receive Ethernet messages or output Ethernet messages;

The encapsulation/decapsulation module is used to perform conversion between Ethernet packets and IB packets.

The buffer module is used to store flight messages, and send the status information of the buffer module to the peer end through the RoCE network.

The IB credit flow control module is used to adjust the message transmission on the IB side.

The QoS module is used to determine the virtual port for transmission between the Ethernet side and the IB side.

The management module is used to manage LIDs, for example, applying for LIDs from the subnet manager and assigning LIDs to nodes.

The message transmission method has been described above, and the communication device that can implement the message transmission method will be described below.

Please refer to FIG. 9, as shown in FIG. 9 is a schematic structural diagram of a communication device provided by the embodiment of the present application. The communication device 90 includes:

The receiving unit 901 is configured to receive a first message from the first network device, the first message header of the first message includes the Internet Protocol IP address of the second network device, and the second network device transmits the message for the first network device The destination network device of the document;

The determining unit 902 is configured to determine the local identifier LID of the second network device according to the IP address of the second network device in combination with a lookup table, and the lookup table includes an association relationship between the IP address and the LID;

The encapsulation unit 903 is configured to strip off the first packet header of the first packet and encapsulate the second packet header to obtain the second packet, the second packet header includes a local routing header, and the local routing header includes a second The LID of the network device;

A sending unit 904, configured to send the second packet according to the LID of the second network device.

Optionally, the receiving unit 901 is specifically used for:

Receive the first message in a large-capacity buffer according to the credit flow control mechanism;

The status information of the large-capacity buffer is fed back to the first network device by suspending the message, so that the first network device adjusts message transmission.

Optionally, the sending unit 904 is also used to:

Apply for a LID from the subnet manager according to the route change, and the route change instructs the first network device to join the network;

The receiving unit 901 is also used for:

receiving the LID of the first network device;

The obtaining unit 905 is also used for:

Obtain a response message from the second network device, where the response message includes the LID and IP address of the second network device;

The communication device further includes an updating unit 906, and the updating unit 906 is specifically used for:

The lookup table is updated according to the IP address and LID of the first network device, and the IP address and LID of the second network device.

Optionally, the sending unit 904 is also used to:

Receive an address resolution protocol ARP message from the first network device, where the ARP message includes the IP address of the first network device and the IP address of the second network device;

Apply for the LID of the first network device from the subnet manager according to the ARP message.

Optionally, the update unit 906 is also used to:

The invariant cyclic redundancy check code ICRC and the variable cyclic redundancy check code VCRC of the second packet are updated.

Optionally, the first message is an Ethernet message, and the second message is an IB message.

Optionally, the Ethernet packet includes an Ethernet header, an IP header, a UDP header, an IB transmission header, an IB payload, an ICRC, and a CRC.

Optionally, the IB packet includes a local routing header, an IB transport header, an IB payload, an ICRC, and a VCRC.

Please refer to FIG. 10 , as shown in FIG. 10 , it is a schematic diagram of another structure of the communication device provided by the embodiment of the present application. The communication device 100 includes:

The receiving unit 1001 is configured to receive a third message from the first network device, the third message header of the third message includes a local routing header, and the local routing header includes a local identifier LID of the second network device, and the second network The device is a destination network device for transmitting packets by the first network device;

The determining unit 1002 is configured to determine the Internet protocol IP address of the second network device according to the LID of the second network device in combination with a lookup table, and the lookup table includes an association relationship between the IP address and the LID;

The encapsulation unit 1003 is configured to strip off the third packet header of the third packet, and encapsulate the fourth packet header to obtain the fourth packet, where the fourth packet header includes the IP address of the second network device;

A sending unit 1004, configured to send a fourth packet according to the IP address of the second network device.

Optionally, the communication device 100 further includes an obtaining unit 1005, and the obtaining unit 1005 is specifically used for:

Obtain the QPN of the first network device and the QPN of the second network device according to the link establishment message;

Acquire the User Datagram Protocol UDP port number of the second network device according to the QPN of the first network device and the QPN of the second network device, the lookup table also includes the association relationship between QPN, UDP port number, IP address and LID, the fourth message The header also includes the MAC address of the media access control layer of the second network device and the UDP port number, and the MAC address of the second network device is obtained by broadcasting the IP address of the second network device.

Optionally, the sending unit 1004 is specifically used for:

Send the fourth message in a large-capacity buffer according to the credit flow control mechanism and the IP address of the second network device;

The status information of the large-capacity buffer is fed back to the second network device by suspending the message, so that the second network device adjusts message transmission.

Optionally, the sending unit 1004 is also used to:

Apply for a LID from the subnet manager according to the route change, and the route change instructs the second network device to join the network;

The receiving unit 1001 is also used for:

receiving the LID of the second network device;

The acquiring unit 1005 is also used for:

Obtain a response message from the first network device, where the response message includes the LID and IP address of the first network device;

The communication device 100 also includes an updating unit 1006, and the updating unit 1006 is specifically used for:

The lookup table is updated according to the IP address and LID of the first network device, and the IP address and LID of the second network device.

Optionally, the sending unit 1004 is also used to:

Receive an address resolution protocol ARP message from the second network device, where the ARP message includes the IP address of the first network device and the IP address of the second network device;

Apply for the LID of the second network device from the subnet manager according to the ARP message.

Optionally, the updating unit 1006 is also used for:

The invariant cyclic redundancy check ICRC and the cyclic redundancy check CRC of the second packet are updated.

Optionally, the first message is an Ethernet message, and the second message is an IB message.

Optionally, the Ethernet packet includes an Ethernet header, an IP header, a UDP header, an IB transmission header, an IB payload, an ICRC, and a CRC.

Optionally, the IB packet includes a local routing header, an IB transport header, an IB payload, an ICRC, and a VCRC.

FIG. 11 is a schematic diagram of a possible logical structure of a communication device 110 provided by an embodiment of the present application. The communication device 110 includes: a processor 1101 , a communication interface 1102 , a storage system 1103 and a bus 1104 . The processor 1101 , the communication interface 1102 and the storage system 1103 are connected to each other through a bus 1104 . In the embodiment of the present application, the processor 1101 is used to control and manage the actions of the communication device 110, for example, the processor 1101 is used to execute the steps performed by the gateway in the method embodiment in FIG. 4 . The communication interface 1102 is used to support the communication device 110 to communicate. The storage system 1103 is configured to store program codes and data of the communication device 110 .

Wherein, the processor 1101 may be a central processing unit, a general purpose processor, a digital signal processor, an application specific integrated circuit, a field programmable gate array or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It can implement or execute the various illustrative logical blocks, modules and circuits described in connection with the present disclosure. The processor 1101 may also be a combination that implements computing functions, for example, a combination of one or more microprocessors, a combination of a digital signal processor and a microprocessor, and the like. The bus 1104 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (Extended Industry Standard Architecture, EISA) bus, etc. The bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one thick line is used in FIG. 11 , but it does not mean that there is only one bus or one type of bus.

The receiving unit 901 and the sending unit 904 in the communication device 90 are equivalent to the communication interface 1102 in the communication device 110, and the determination unit 902, encapsulation unit 903, acquisition unit 905 and update unit 906 in the communication device 90 are equivalent to the communication device 110. Processor 1101.

The communication device 110 in this embodiment may correspond to the gateway in the above-mentioned embodiment of the method in FIG. Various steps are not repeated here for the sake of brevity.

FIG. 12 is a schematic diagram of a possible logical structure of a communication device 120 provided by an embodiment of the present application. The communication device 120 includes: a processor 1201 , a communication interface 1202 , a storage system 1203 and a bus 1204 . The processor 1201 , the communication interface 1202 and the storage system 1203 are connected to each other through the bus 1204 . In the embodiment of the present application, the processor 1201 is used to control and manage the actions of the communication device 120, for example, the processor 1201 is used to execute the steps performed by the gateway in the method embodiment in FIG. 5 . The communication interface 1202 is used to support the communication device 120 to communicate. The storage system 1203 is used for storing program codes and data of the communication device 120 .

Wherein, the processor 1201 may be a central processing unit, a general processor, a digital signal processor, an application specific integrated circuit, a field programmable gate array or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It can implement or execute the various illustrative logical blocks, modules and circuits described in connection with the present disclosure. The processor 1201 may also be a combination that implements computing functions, for example, a combination of one or more microprocessors, a combination of a digital signal processor and a microprocessor, and the like. The bus 1204 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (Extended Industry Standard Architecture, EISA) bus, etc. The bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one thick line is used in FIG. 12 , but it does not mean that there is only one bus or one type of bus.

The receiving unit 1001 and the sending unit 1004 in the communication device 100 are equivalent to the communication interface 1202 in the communication device 120, and the determination unit 1002, encapsulation unit 1003, acquisition unit 1005 and update unit 1006 in the communication device 100 are equivalent to the communication interface 1202 in the communication device 120. Processor 1201.

The communication device 120 in this embodiment may correspond to the gateway in the above-mentioned method embodiment in FIG. Various steps are not repeated here for the sake of brevity.

In another embodiment of the present application, a computer-readable storage medium is also provided, in which computer-executable instructions are stored, and when the processor of the device executes the computer-executable instructions, the device executes the above-mentioned method in Figure 4 Steps of the message transmission method executed by the gateway device in the embodiment.

In another embodiment of the present application, a computer-readable storage medium is also provided. Computer-readable storage medium stores computer-executable instructions. When the processor of the device executes the computer-executable instructions, the device executes the above-mentioned method in FIG. 5 Steps of the message transmission method performed by the gateway in the embodiment.

In another embodiment of the present application, a computer program product is also provided, the computer program product includes computer-executable instructions stored in a computer-readable storage medium; when the processor of the device executes the computer-executable instructions , the device executes the steps of the packet transmission method executed by the gateway in the method embodiment in FIG. 4 above.

In another embodiment of the present application, a computer program product is also provided, the computer program product includes computer-executable instructions stored in a computer-readable storage medium; when the processor of the device executes the computer-executable instructions , the device executes the steps of the packet transmission method executed by the gateway in the method embodiment in FIG. 5 above.

Those skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or May be integrated into another system, or some features may be ignored, or not implemented. In another point, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

If the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or part of the contribution to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), magnetic disk or optical disc, etc., which can store program codes. .

Claims (38)

1. A message transmission method, characterized in that, comprising:

   The gateway receives the first message from the first network device, the first message header of the first message includes the Internet Protocol IP address of the second network device, and the second network device transmits the message for the first network device the destination network device;

   The gateway determines the local identifier LID of the second network device according to the IP address of the second network device in combination with a lookup table, and the lookup table includes an association relationship between the IP address and the LID;

   The gateway strips the first packet header of the first packet and encapsulates the second packet header to obtain the second packet, the second packet header includes a local routing header, and the local routing header includes the the LID of the second network device;

   The gateway sends the second packet according to the LID of the second network device.
2. The message transmission method according to claim 1, wherein the gateway receiving the first message from the first network device comprises:

   The gateway receives the first message in a large-capacity buffer according to a credit flow control mechanism;

   The gateway feeds back the status information of the large-capacity buffer to the first network device by suspending the message, so that the first network device adjusts message transmission.
3. The message transmission method according to any one of claims 1-2, wherein before the gateway receives the first message from the first network device, the method further comprises:

   The gateway applies for a LID from the subnet manager according to a route change, and the route change instructs the first network device to join the network;

   the gateway receives the LID of the first network device;

   The gateway acquires a response message from the second network device, where the response message includes the LID and IP address of the second network device;

   The gateway updates the lookup table according to the IP address and LID of the first network device and the IP address and LID of the second network device.
4. The message transmission method according to claim 3, wherein the gateway applying for a LID from the subnet manager according to the routing change comprises:

   The gateway receives an Address Resolution Protocol ARP message from the first network device, where the ARP message includes the IP address of the first network device and the IP address of the second network device;

   The gateway applies for the LID of the first network device from the subnet manager according to the ARP message.
5. The message transmission method according to any one of claims 1-4, wherein the gateway strips the first message header of the first message and encapsulates the second message header to obtain the second message header. After the text, the method also includes:

   The gateway updates the invariant cyclic redundancy check code ICRC and the variable cyclic redundancy check code VCRC of the second message.
6. The message transmission method according to any one of claims 1-5, characterized in that, the first message is an Ethernet message, and the second message is an IB message.
7. The message transmission method according to claim 6, wherein the Ethernet message includes an Ethernet header, an IP header, a UDP header, an IB transmission header, an IB payload, an ICRC, and a cyclic redundancy check code (CRC).
8. The message transmission method according to claim 6, wherein the IB message includes the local routing header, the IB transmission header, the IB payload, ICRC and VCRC.
9. A message transmission method, characterized in that, comprising:

   The gateway receives the third message from the first network device, the third message header of the third message includes a local routing header, and the local routing header includes a local identifier LID of the second network device, and the second The network device is the destination network device for transmitting the message by the first network device;

   The gateway determines the Internet Protocol IP address of the second network device according to the LID of the second network device in combination with a lookup table, and the lookup table includes an association relationship between the IP address and the LID;

   The gateway strips the third packet header of the third packet, and encapsulates the fourth packet header to obtain the fourth packet, and the fourth packet header includes the IP address of the second network device;

   The gateway sends the fourth packet according to the IP address of the second network device.
10. The message transmission method according to claim 9, wherein the method further comprises:

    The gateway obtains the queue pair sequence number QPN of the first network device and the QPN of the second network device according to the link establishment message;

    The gateway acquires the User Datagram Protocol UDP port number of the second network device according to the QPN of the first network device and the QPN of the second network device, and the lookup table further includes QPN, UDP port number, IP The association relationship between the address and the LID, the fourth packet header also includes the MAC address of the media access control layer and the UDP port number of the second network device, and the MAC address of the second network device is based on the second network The IP address of the device is obtained by broadcasting.
11. The message transmission method according to claim 9 or 10, wherein the gateway sending the fourth message according to the IP address of the second network device includes:

    The gateway sends the fourth message in a large-capacity buffer according to the credit flow control mechanism and the IP address of the second network device;

    The gateway feeds back the state information of the large-capacity buffer to the second network device by suspending the message, so that the second network device adjusts message transmission.
12. The message transmission method according to any one of claims 9-11, wherein before the gateway receives the third message from the first network device, the method further comprises:

    The gateway applies for a LID from the subnet manager according to a route change, and the route change instructs the second network device to join the network;

    the gateway receives the LID of the second network device;

    The gateway acquires a response message from the first network device, where the response message includes the LID and IP address of the first network device;

    The gateway updates the lookup table according to the IP address and LID of the first network device and the IP address and LID of the second network device.
13. The message transmission method according to claim 12, wherein the gateway applying for the LID from the subnet manager according to the route change comprises:

    The gateway receives an Address Resolution Protocol ARP message from the second network device, where the ARP message includes the IP address of the first network device and the IP address of the second network device;

    The gateway applies for the LID of the second network device from the subnet manager according to the ARP message.
14. The message transmission method according to any one of claims 9-13, wherein the gateway strips the first message header of the first message and encapsulates the second message header to obtain the second message header. After the text, the method also includes:

    The gateway updates the invariant cyclic redundancy check ICRC and the cyclic redundancy check CRC of the second message.
15. The message transmission method according to any one of claims 9-14, wherein the first message is an Ethernet message, and the second message is an IB message.
16. The message transmission method according to claim 15, wherein the Ethernet message includes an Ethernet header, an IP header, a UDP header, an IB transmission header, an IB payload, an ICRC and a CRC.
17. The message transmission method according to claim 15, wherein the IB message includes the local routing header, the IB transmission header, the IB payload, ICRC and a variable cyclic redundancy check code VCRC.
18. A communication device, characterized by comprising:

    A receiving unit, configured to receive a first message from a first network device, the first message header of the first message includes an Internet Protocol IP address of a second network device, and the second network device is the first network device The destination network device for the device to transmit the message;

    A determining unit, configured to determine the local identifier LID of the second network device according to the IP address of the second network device in combination with a lookup table, the lookup table including an association relationship between the IP address and the LID;

    An encapsulation unit, configured to strip off the first packet header of the first packet, and encapsulate the second packet header to obtain the second packet, the second packet header includes a local routing header, and the local routing header including the LID of the second network device;

    A sending unit, configured to send the second packet according to the LID of the second network device.
19. The communication device according to claim 18, wherein the receiving unit is specifically used for:

    receiving the first message in a large-capacity buffer according to a credit flow control mechanism;

    The status information of the large-capacity buffer is fed back to the first network device by suspending the message, so that the first network device adjusts message transmission.
20. The communication device according to any one of claims 18-19, wherein the sending unit is further configured to:

    Applying for a LID from the subnet manager according to a route change, the route change instructing the first network device to join the network;

    The receiving unit is also used for:

    receiving the LID of the first network device;

    The acquisition unit is also used for:

    Obtain a response message from the second network device, where the response message includes the LID and IP address of the second network device;

    The communication device also includes an updating unit, and the updating unit is specifically used for:

    updating the lookup table according to the IP address and LID of the first network device and the IP address and LID of the second network device.
21. The communication device according to claim 20, wherein the sending unit is further used for:

    receiving an Address Resolution Protocol ARP message from the first network device, where the ARP message includes the IP address of the first network device and the IP address of the second network device;

    Applying for the LID of the first network device from the subnet manager according to the ARP message.
22. The communication device according to any one of claims 18-21, wherein the updating unit is further configured to:

    Updating the invariant cyclic redundancy check code ICRC and the variable cyclic redundancy check code VCRC of the second packet.
23. The communication device according to any one of claims 18-22, wherein the first message is an Ethernet message, and the second message is an IB message.
24. The communication device according to claim 23, wherein the Ethernet message includes an Ethernet header, an IP header, a UDP header, an IB transmission header, an IB payload, an ICRC, and a cyclic redundancy check code (CRC).
25. The communication device according to claim 23, wherein the IB message includes the local routing header, IB transmission header, IB payload, ICRC and VCRC.
26. A communication device, characterized by comprising:

    a receiving unit, configured to receive a third message from the first network device, the third message header of the third message includes a local routing header, and the local routing header includes a local identifier LID of the second network device, The second network device is a destination network device for transmitting packets by the first network device;

    A determining unit, configured to determine the Internet Protocol IP address of the second network device according to the LID of the second network device in combination with a lookup table, the lookup table including an association relationship between the IP address and the LID;

    An encapsulation unit, configured to strip off the third packet header of the third packet, and encapsulate the fourth packet header to obtain the fourth packet, the fourth packet header including the IP address of the second network device ;

    a sending unit, configured to send the fourth packet according to the IP address of the second network device.
27. The communication device according to claim 26, wherein the communication device further comprises an acquisition unit, and the acquisition unit is specifically used for:

    Obtain the queue pair sequence number QPN of the first network device and the QPN of the second network device according to the link establishment message;

    Obtain the User Datagram Protocol UDP port number of the second network device according to the QPN of the first network device and the QPN of the second network device, and the lookup table also includes QPN, UDP port number, IP address and LID The fourth packet header also includes the MAC address of the media access control layer and the UDP port number of the second network device, and the MAC address of the second network device is based on the IP address of the second network device. The address is obtained by broadcasting.
28. The communication device according to claim 26 or 27, wherein the sending unit is specifically used for:

    sending the fourth message in a large-capacity buffer according to the credit flow control mechanism and the IP address of the second network device;

    The status information of the large-capacity buffer is fed back to the second network device by suspending the message, so that the second network device adjusts message transmission.
29. The communication device according to any one of claims 26-28, wherein the sending unit is further configured to:

    Applying for a LID from the subnet manager according to a route change, the route change instructing the second network device to join the network;

    The receiving unit is also used for:

    receiving the LID of the second network device;

    The acquisition unit is also used for:

    Obtain a response message from the first network device, where the response message includes the LID and IP address of the first network device;

    The communication device also includes an updating unit, and the updating unit is specifically used for:

    updating the lookup table according to the IP address and LID of the first network device and the IP address and LID of the second network device.
30. The communication device according to claim 29, wherein the sending unit is further used for:

    receiving an Address Resolution Protocol ARP message from the second network device, where the ARP message includes the IP address of the first network device and the IP address of the second network device;

    Applying for the LID of the second network device from the subnet manager according to the ARP message.
31. The communication device according to any one of claims 26-30, wherein the update unit is further configured to:

    Updating the invariant cyclic redundancy check ICRC and the cyclic redundancy check CRC of the second packet.
32. The communication device according to any one of claims 26-31, wherein the first message is an Ethernet message, and the second message is an IB message.
33. The communication device according to claim 32, wherein the Ethernet packet includes an Ethernet header, an IP header, a UDP header, an IB transmission header, an IB payload, an ICRC and a CRC.
34. The communication device according to claim 32, wherein the IB message includes the local routing header, IB transmission header, IB payload, ICRC and variable cyclic redundancy check code VCRC.
35. A communication device, characterized in that it includes: a processor and a memory,

    The processor is configured to execute instructions stored in the memory, so that the communication device executes the method according to any one of claims 1-8.
36. A communication device, characterized in that it includes: a processor and a memory,

    The processor is configured to execute instructions stored in the memory, so that the communication device executes the method according to any one of claims 9 to 17.
37. A computer-readable storage medium, characterized in that a computer program is stored in the computer-readable storage medium, and when the computer program is run on the computer, the computer executes the any one of the methods described.
38. A computer program product, characterized in that, when the computer program product is executed on a computer, the computer executes the method according to any one of claims 1 to 17.

Images