VRB-based multi-protocol improved intercommunication fusion method, apparatus and device, and storage medium

Abstract

The embodiment of the invention provides a VRB multi-protocol-based improved intercommunication fusion method, device and equipment and a storage medium, and the method comprises the steps: receiving a first data packet sent by first RDMA equipment, the first data packet comprises first data header information and first data content of an RDMA protocol, and the first data header information comprises an IP address of second articulated naturality web equipment; obtaining first key information used for identifying the first RDMA equipment and second key information used for identifying the second articulated naturality web equipment; removing first data header information in the first data packet to obtain a second data packet; the second data is processed to obtain a third data packet, the third data packet comprises first articulated naturality web protocol information and first data content, and the first articulated naturality web protocol information comprises first key information and second key information; and sending a third data packet to the second articulated naturality web device based on the first articulated naturality web protocol information. Therefore, according to the embodiment of the invention, communication between devices with different communication protocols in a large model cluster can be realized.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a method, apparatus, device, and storage medium for improved intercommunication and fusion based on VRB multi-protocol. Background Art

[0002] A large model cluster is a system consisting of multiple computing nodes, each with powerful computing capabilities. These nodes are interconnected via a high-speed network to collaboratively complete complex computing tasks. Large model clusters are primarily used to support large-scale deep learning model training and inference, and can be applied to fields such as artificial intelligence, big data analysis, and scientific computing.

[0003] Currently, large model clusters can leverage distributed computing technology to break down training tasks into multiple subtasks and assign them to different computing nodes for parallel processing. Communication is essential for distributed training within large model clusters. However, currently, each node in a large model cluster only supports communication based on a single protocol, making it impossible to deploy devices using different communication protocols within the cluster.

[0004] It can be seen from this that the existing technology cannot achieve communication between devices with different communication protocols in a large model cluster. Summary of the Invention

[0005] The embodiments of the present application provide a method, apparatus, device and storage medium for improved intercommunication fusion based on VRB multi-protocol to achieve communication between devices of different communication protocols in a large model cluster.

[0006] In a first aspect, an embodiment of the present application provides an improved intercommunication and fusion method based on VRB multi-protocol, applied to a first visual network device, the method comprising:

[0007] Receive a first data packet sent by a first RDMA device, wherein the first data packet includes first data header information and first data content of an RDMA protocol, the first data header information includes an IP address of a second visual network device, and RDMA stands for remote direct memory access;

[0008] Acquire first key information for identifying the first RDMA device;

[0009] Acquire second key information for identifying the second visual network device;

[0010] removing the first data header information from the first data packet to obtain a second data packet;

[0011] encapsulating a visual networking protocol for the second data packet, and processing the second data packet based on the first key information and the second key information to obtain a third data packet, wherein the third data packet includes the first visual networking protocol information and the first data content, and the first visual networking protocol information includes the first key information and the second key information;

[0012] Based on the first visual networking protocol information, the third data packet is sent to the second visual networking device, so that the second visual networking device transmits the first data content to the first large model deployed on the second visual networking device according to the first key information and the second key information.

[0013] In a second aspect, an embodiment of the present application provides an improved intercommunication and fusion method based on VRB multi-protocol, applied to a second visual network device, the method comprising:

[0014] Receiving a third data packet sent by the first visual networking device based on the first visual networking protocol information, wherein the third data packet includes the first visual networking protocol information and the first data content, and the first visual networking protocol information includes first key information for identifying the first RDMA device and second key information for identifying the second visual networking device;

[0015] According to the first key information and the second key information, the first data content is transmitted to the first large model deployed on the second visual network device.

[0016] In a third aspect, an embodiment of the present application provides an improved intercommunication and fusion device based on VRB multi-protocol, applied to a first visual network device, the device comprising:

[0017] A first receiving module is configured to receive a first data packet sent by a first RDMA device, wherein the first data packet includes first data header information and first data content of an RDMA protocol, the first data header information includes an IP address of a second visual network device, and RDMA stands for Remote Direct Memory Access;

[0018] A first acquisition module, configured to acquire first key information for identifying the first RDMA device;

[0019] A second acquisition module, configured to acquire second key information for identifying the second visual network device;

[0020] a removal module, configured to remove the first data header information from the first data packet to obtain a second data packet;

[0021] a first processing module, configured to encapsulate a visual networking protocol for the second data packet, and process the second data packet based on the first key information and the second key information to obtain a third data packet, wherein the third data packet includes the first visual networking protocol information and the first data content, and the first visual networking protocol information includes the first key information and the second key information;

[0022] The first sending module is used to send the third data packet to the second visual networking device based on the first visual networking protocol information, so that the second visual networking device transmits the first data content to the first large model deployed on the second visual networking device according to the first key information and the second key information.

[0023] In a fourth aspect, an embodiment of the present application provides an improved intercommunication and fusion device based on VRB multi-protocol, applied to a second visual network device, the device comprising:

[0024] a second receiving module, configured to receive a third data packet sent by the first visual networking device based on the first visual networking protocol information, wherein the third data packet includes the first visual networking protocol information and the first data content, and the first visual networking protocol information includes first key information for identifying the first RDMA device and second key information for identifying the second visual networking device;

[0025] The second sending module is used to transmit the first data content to the first large model deployed on the second visual network device according to the first key information and the second key information.

[0026] In a fifth aspect, an embodiment of the present application provides an electronic device, including:

[0027] one or more processors; and

[0028] One or more machine-readable media having instructions stored thereon, when executed by the one or more processors, cause the electronic device to execute the improved intercommunication fusion method based on VRB multi-protocol described in the first aspect above, or execute the improved intercommunication fusion method based on VRB multi-protocol described in the second aspect above.

[0029] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program that enables a processor to execute the improved intercommunication and fusion method based on VRB multi-protocol described in the first aspect above, or to execute the improved intercommunication and fusion method based on VRB multi-protocol described in the second aspect above.

[0030] In some embodiments of the present application, a first visual networking device can receive a first data packet sent by a first RDMA device, wherein the first data packet includes first data header information and first data content of the RDMA protocol, and the first data header information includes the IP address of the second visual networking device; and obtain first key information for identifying the first RDMA device and second key information for identifying the second visual networking device, thereby removing the first data header information in the first data packet to obtain a second data packet, and then encapsulating the visual networking protocol for the second data packet, and processing the second data packet based on the first key information and the second key information to obtain a third data packet, wherein the third data packet includes first visual networking protocol information and first data content, and the first visual networking protocol information includes the first key information and the second key information; and then based on the first visual networking protocol information, send the third data packet to the second visual networking device, so that the second visual networking device transmits the first data content to the first large model deployed on the second visual networking device according to the first key information and the second key information.

[0031] It can be seen that in some embodiments of the present application, visual networking devices and RDMA devices can be deployed in a large model cluster, and in the process of transmitting the first data packet from the first RDMA device to the second visual networking device, the first visual networking device can remove the original first data header information of the RDMA protocol in the first data packet, and encapsulate the visual networking protocol information including the first key information and the second key information for it, thereby obtaining the above-mentioned third data packet. After the second visual networking device receives the third data packet, it can transmit the first data content to the first large model for processing based on the first key information and the second key information in the third data packet.

[0032] As can be seen, in some embodiments of the present application, by encapsulating the visual networking protocol information including the first key information and the second key information in an RDMA protocol data packet, a first visual networking device can achieve data transmission between the first RDMA device and the second visual networking device in a large model cluster, thereby achieving the interoperability and integration of multiple protocols in the large model. Furthermore, by removing the original RDMA protocol first data header information in the first data packet, network transmission pressure can be reduced, thereby improving network transmission performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a flowchart of the steps of the VRB multi-protocol improved intercommunication and integration method of the present application;

[0034] Figure 2 This is a flowchart of another improved VRB multi-protocol intercommunication and integration method of the present application;

[0035] Figure 3This is a schematic diagram of an architecture applicable to the VRB multi-protocol improved intercommunication and integration method according to an embodiment of the present application;

[0036] Figure 4 Schematic diagram of the transmission process of the first data content in the VRB multi-protocol improved intercommunication and integration method according to an embodiment of the present application;

[0037] Figure 5 1 is a schematic diagram of a transmission process of the second data content in the VRB multi-protocol improved intercommunication and integration method according to an embodiment of the present application;

[0038] Figure 6 2 is a schematic diagram of a transmission process of the third data content in the VRB multi-protocol improved intercommunication and integration method according to an embodiment of the present application;

[0039] Figure 7 2 is a schematic diagram of a transmission process of the fourth data content in the VRB multi-protocol improved intercommunication and integration method according to an embodiment of the present application;

[0040] Figure 8 Schematic diagram of a specific implementation of the VRB multi-protocol improved intercommunication and integration method according to an embodiment of the present application;

[0041] Figure 9 This is a structural block diagram of an improved VRB multi-protocol intercommunication and fusion device based on the present application;

[0042] Figure 10 This is a structural block diagram of another improved VRB multi-protocol intercommunication and fusion device in this application. DETAILED DESCRIPTION

[0043] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0044] In the prior art, each node in a large model cluster only supports communication based on one communication protocol, which makes it impossible to deploy devices with different communication protocols in the large model cluster.

[0045] In some embodiments of the present application, visual networking devices and RDMA devices can be deployed in a large model cluster, and in the process of transmitting a first data packet from a first RDMA device to a second visual networking device, the first visual networking device can remove the original first data header information of the RDMA protocol in the first data packet, and encapsulate the visual networking protocol information including the first key information and the second key information for it, thereby obtaining the above-mentioned third data packet, and after the second visual networking device receives the third data packet, it can transmit the first data content to the first large model for processing based on the first key information and the second key information in the third data packet. Therefore, some embodiments of the present application can realize communication between devices with different communication protocols in a large model cluster.

[0046] The following is a detailed description of the VRB multi-protocol improved intercommunication and integration method according to the embodiment of the present application. Figure 1 As shown, an embodiment of the present application provides an improved intercommunication and fusion method based on VRB multi-protocol; wherein, VRB is a remote direct memory access (RDMA) technology system implemented by the V2V communication protocol; that is, the VRB protocol is a remote data reading protocol provided by the visual network to support the transmission of large data in the intelligent computing center.

[0047] The VRB multi-protocol improved intercommunication fusion method can be applied to the first visual network device, such as Figure 1 As shown, the method may include the following steps 101 to 106:

[0048] Step 101: Receive a first data packet sent by a first RDMA device.

[0049] The first data packet includes first data header information and first data content of a Remote Direct Memory Access (RDMA) protocol, and the first data header information includes an IP address of a second visual network device.

[0050] Optionally, the first datagram header information may include an Internet Protocol (IP) header and a User Datagram Protocol (UDP) header.

[0051] The IP header includes the source IP address and the destination IP address; for example, when the first data packet needs to be sent by the first RDMA device to the first visual networking device, and then the first visual networking device sends the first data content in the first data packet to the second visual networking device, the source IP address in the IP header of the first data packet is the IP address of the first RDMA device, and the destination IP address is the IP address of the second visual networking device.

[0052] The UDP header is the part of the UDP protocol used to transmit data packets. It includes some key information to ensure the correct transmission and processing of data packets.

[0053] Optionally, the first data packet also includes a source physical address and a destination physical address. For example, if the first data packet needs to be sent from the first RDMA device to the first visual networking device, and then the first visual networking device transmits the first data content in the first data packet to the second visual networking device, the source physical address in the first data packet is the physical address of the first RDMA device, and the destination physical address is the physical address of the first visual networking device. The physical address may be a Media Access Control (MAC) address.

[0054] In addition, the first data content can be understood as data content that needs to be transmitted to the second visual networking device for processing by the first large model deployed on the second visual networking device; that is, it can be understood as the first data content including the input content of the first large model. As can be seen, in some embodiments of the present application, when the first data content of the first RDMA device needs to be processed by the first large model deployed on the second visual networking device, the first RDMA device can first transmit the first data content to the first visual networking device, and then the first visual networking device can transmit it to the second visual networking device.

[0055] Step 102: Acquire first key information for identifying the first RDMA device.

[0056] The first key information is used to identify address information of the first RDMA device; optionally, the first key information includes key information generated according to the IP address of the first RDMA device.

[0057] Step 103: Obtain second key information for identifying the second visual network device.

[0058] The second key information is used to identify the address information of the second visual networking device; optionally, the second key information includes key information generated according to the IP address of the second visual networking device.

[0059] Step 104: Remove the first data header information from the first data packet to obtain a second data packet.

[0060] As can be seen from step 104, after receiving the first data packet, the first visual networking device can remove the first data header information of the RDMA protocol in the first data packet.

[0061] Step 105: Encapsulate the visual networking protocol for the second data packet, and process the second data packet based on the first key information and the second key information to obtain a third data packet.

[0062] Among them, the third data packet includes the first visual network protocol information and the first data content, and the first visual network protocol information includes the first key information and the second key information.

[0063] As can be seen from step 105, after removing the first data header information from the first data packet, the first visual networking device can further encapsulate the remaining portion with a visual networking protocol based on the first key information and the second key information. In this way, the first visual networking device can send the second data packet to the second visual networking device based on the first visual networking protocol information.

[0064] Optionally, the first visual networking protocol information also includes a source visual networking number and a destination visual networking number; for example, when the third data packet needs to be sent from the first visual networking device to the second visual networking device, the source visual networking number in the first visual networking protocol information is the visual networking number of the first visual networking device, and the destination visual networking number is the visual networking number of the second visual networking device.

[0065] Optionally, the third data packet further includes a source physical address and a destination physical address. For example, when the third data packet needs to be sent from the first visual networking device to the second visual networking device, the source physical address in the second data packet is the physical address of the first visual networking device, and the destination physical address is the physical address of the second visual networking device. The physical address may be a Media Access Control (MAC) address.

[0066] Step 106: Based on the first visual networking protocol information, send the third data packet to the second visual networking device, so that the second visual networking device transmits the first data content to the first large model deployed on the second visual networking device according to the first key information and the second key information.

[0067] It can be seen from the above steps 104 to 105 that after the first visual networking device receives the first data packet of the RDMA protocol sent by the first RDMA device, although the original first data header information of the RDMA protocol is removed, the first visual networking protocol information carrying the above-mentioned first key information and second key information will be added to the remaining part. In this way, the third data packet of the visual networking protocol obtained contains the first key information for identifying the first RDMA device and the second key information for identifying the second visual networking device, so that after the second visual networking device receives the second data packet, it does not need to convert the relevant information of the RDMA protocol based on the first visual networking protocol information, and can directly transmit the first data content to the first large model deployed on the second visual networking device based on the first key information and the second key information.

[0068] Exemplarily, when the first datagram header information includes an IP header and a UDP header, the IP header includes a source IP address (i.e., the IP address of the RDMA device) and a destination IP address (i.e., the IP address of the second visual networking device), and the UDP header includes verification information, after the first visual networking device receives the first data packet including the first datagram header information and the first data content, it can remove the IP header and the UDP header, and encapsulate the first visual networking protocol information including the first key information and the second key information based on the remaining portion to obtain a third data packet, thereby sending the third data packet to the second visual networking device. In this way, since the first visual networking protocol information includes verification information and also contains the first key information for identifying the first RDMA device and the second key information for identifying the second visual networking device, after the second visual networking device receives the third data packet, it does not need to convert the destination visual networking number (i.e., the visual networking number of the second visual networking device) in the first visual networking protocol information into an IP address, and can directly use the first key information, the second key information, and the verification information in the first visual networking protocol information to transmit the first data content to the first large model deployed on the second visual networking device.

[0069] Optionally, the first visual networking device can be connected to multiple visual networking devices. In this case, the fourth correspondence between the physical address and port of each visual networking device connected to the first visual networking device can be recorded. In this way, when the third data packet includes the physical address of the second visual networking device, the first visual networking device can determine the corresponding port based on the physical address of the second visual networking device in the third data packet and the fourth correspondence, and then send the third data packet to the second visual networking device through the port.

[0070] It can be seen from the above steps 101 to 106 that in some embodiments of the present application, the first visual networking device can receive a first data packet sent by the first RDMA device, wherein the first data packet includes the first data header information and the first data content of the RDMA protocol, and the first data header information includes the IP address of the second visual networking device; and obtain the first key information for identifying the first RDMA device and the second key information for identifying the second visual networking device, thereby removing the first data header information in the first data packet to obtain a second data packet, and then encapsulating the visual networking protocol for the second data packet, and processing the second data packet based on the first key information and the second key information to obtain a third data packet, wherein the third data packet includes the first visual networking protocol information and the first data content, and the first visual networking protocol information includes the first key information and the second key information; and then based on the first visual networking protocol information, send the third data packet to the second visual networking device, so that the second visual networking device transmits the first data content to the first large model deployed on the second visual networking device according to the first key information and the second key information.

[0071] It can be seen that in some embodiments of the present application, visual networking devices and RDMA devices can be deployed in a large model cluster, and in the process of transmitting the first data packet from the first RDMA device to the second visual networking device, the first visual networking device can remove the original first data header information of the RDMA protocol in the first data packet, and encapsulate the visual networking protocol information including the first key information and the second key information for it, thereby obtaining the above-mentioned third data packet. After the second visual networking device receives the third data packet, it can transmit the first data content to the first large model for processing based on the first key information and the second key information in the third data packet.

[0072] As can be seen, in some embodiments of the present application, by encapsulating the visual networking protocol information including the first key information and the second key information in an RDMA protocol data packet, a first visual networking device can achieve data transmission between the first RDMA device and the second visual networking device in a large model cluster, thereby achieving the interoperability and integration of multiple protocols in the large model. Furthermore, by removing the original RDMA protocol first data header information in the first data packet, network transmission pressure can be reduced, thereby improving network transmission performance.

[0073] In some embodiments of the present application, in step 102, obtaining the first key information for identifying the first RDMA device includes the following step A-1 or A-2:

[0074] Step A-1: ​​when the first data header information further includes the IP address of the first RDMA device, obtaining the first key information corresponding to the IP address of the first RDMA device according to a stored first correspondence between the IP address of the RDMA device and the key information;

[0075] It can be seen from step A-1 that the first correspondence between the IP addresses and key information of each RDMA device connected to the first visual networking device can be pre-stored, so that after the first visual networking device receives the above-mentioned first data packet, it can obtain the first key information corresponding to the IP address of the first RDMA device based on the first correspondence.

[0076] Step A-2: When the first data packet further includes the physical address of the first RDMA device, the first key information corresponding to the physical address of the first RDMA device is obtained according to a stored second correspondence between the physical address of the RDMA device and the key information.

[0077] It can be seen from step A-2 that the second correspondence between the physical addresses and key information of each RDMA device connected to the first visual networking device can be pre-stored, so that after the first visual networking device receives the above-mentioned first data packet, it can obtain the first key information corresponding to the physical address of the first RDMA device based on the second correspondence.

[0078] In some embodiments of the present application, in the above step 103, the obtaining of the second key information for identifying the second visual network device includes the following step B-1:

[0079] Step B-1: According to the third correspondence between the IP address of the visual networking device and the key information that has been stored, obtain the second key information corresponding to the IP address of the second visual networking device.

[0080] It can be seen from step B-1 that the third correspondence between the IP addresses and key information of each visual network device connected to the first visual network device can be pre-stored, so that after the first visual network device receives the above-mentioned first data packet, it can obtain the second key information corresponding to the IP address of the second visual network device based on the third correspondence.

[0081] In some embodiments of the present application, the method further comprises the following steps C-1 to C-2:

[0082] Step C-1: Obtaining the physical address of the second visual networking device according to the IP address of the second visual networking device;

[0083] Step C-2: Encapsulating the physical address of the second visual network device into the third data packet;

[0084] In the above step 106, sending the third data packet to the second visual networking device based on the first visual networking protocol information includes:

[0085] Based on the first visual networking protocol information and the physical address of the second visual networking device, the third data packet is sent to the second visual networking device.

[0086] It can be seen from this that the first data packet includes the IP address of the second visual networking device in the first data packet. In this way, the first visual networking device can obtain the physical address of the second visual networking device based on the IP address, and thus encapsulate the physical address of the second visual networking device into the third data packet. In this way, the first visual networking device can send the third data packet to the second visual networking device based on the first visual networking protocol information and according to the physical address of the second visual networking device.

[0087] Optionally, acquiring the physical address of the second visual networking device according to the IP address of the second visual networking device includes:

[0088] According to the sixth correspondence between the stored physical address and IP address of the visual networking device, the physical address corresponding to the IP address of the second visual networking device is obtained and determined as the physical address of the second visual networking device.

[0089] From this, it can be seen that the first visual networking device can pre-record the sixth correspondence between the physical address and IP address of each visual networking device connected to it. In this way, the first visual networking device can obtain the physical address of the second visual networking device based on the IP address of the second visual networking device and the sixth correspondence.

[0090] In some embodiments of the present application, the method further includes the following steps D-1 to D-5:

[0091] Step D-1: receiving a fourth data packet sent by the second visual networking device based on second visual networking protocol information, wherein the fourth data packet includes the second visual networking protocol information and second data content, and the second visual networking protocol information includes the first key information and the second key information;

[0092] Step D-2: acquiring an IP address corresponding to the first key information according to a stored first correspondence between the IP address of the RDMA device and the key information, and determining the IP address corresponding to the first key information as the IP address of the first RDMA device;

[0093] Step D-3: obtaining the IP address corresponding to the second key information based on the stored third correspondence between the IP address of the visual network device and the key information, and determining the IP address corresponding to the second key information as the IP address of the second visual network device;

[0094] Step D-4: Processing the fourth data packet to obtain a fifth data packet, wherein the fifth data packet includes second data header information of the RDMA protocol and the second data content, and the second data header information includes the IP address of the first RDMA device and the IP address of the second visual network device;

[0095] Step D-5: Based on the second data header information, send the fifth data packet to the first RDMA device.

[0096] Among them, the second data content includes: the content output by the first large model after the first data content is input into the first large model; or, the second data content includes the data content actively sent by the second visual networking device to the first RDMA device (that is, the content that needs to be input into the second large model deployed on the first RDMA device).

[0097] In addition, if the second data content needs to be transmitted from the second visual networking device to the first RDMA device, the second visual networking device can encapsulate the second visual networking protocol information for the second data content based on the first key information used to identify the first RDMA device and the second key information used to identify the second visual networking device, obtain the above-mentioned fourth data packet, and then send the fourth data packet to the first visual networking device based on the second visual networking protocol information, so that the first visual networking device can then send the second data content therein to the first RDMA device.

[0098] It can be seen from the above steps D-1 to D-5 that after the first visual networking device receives the fourth data packet, it can obtain the IP address of the first RDMA device based on the above first correspondence, and obtain the IP address of the second visual networking device based on the above third correspondence, thereby generating the second data packet header information of the RDMA protocol based on the IP address of the first RDMA device and the IP address of the second visual networking device, and encapsulating the second data header information for the second data content to obtain the above fifth data packet. In this way, the first visual networking device can send the fifth data packet to the first RDMA device based on the second data header information.

[0099] Optionally, the second datagram header information may include an IP header and a UDP header. The IP header includes a source IP address and a destination IP address. For example, if the fourth data packet needs to be sent from the second visual networking device to the first visual networking device, and then the first visual networking device sends the second data content in the fourth data packet to the first RDMA device, the source IP address in the IP header of the second datagram header information is the IP address of the second visual networking device, and the destination IP address is the IP address of the first RDMA device.

[0100] Optionally, the fourth data packet also includes a source physical address and a destination physical address; for example, when the fourth data packet needs to be sent by the second visual networking device to the first visual networking device, and then the first visual networking device sends the second data content in the fourth data packet to the first RDMA device, the source physical address in the fourth data packet is the physical address of the second visual networking device, and the destination physical address is the physical address of the first visual networking device.

[0101] In some embodiments of the present application, the method further includes the following steps E-1 to E-2:

[0102] Step E-1: Acquire the physical address of the first RDMA device according to the first key information or the IP address of the first RDMA device;

[0103] Step E-1: Encapsulating the physical address of the first RDMA device into the fifth data packet;

[0104] In the above step D-5, sending the fifth data packet to the first RDMA device based on the second data header information includes:

[0105] The fifth data packet is sent to the first RDMA device based on the second data header information and the physical address of the first RDMA device.

[0106] It can be seen from this that the above-mentioned fourth data packet includes the first key information of the first RDMA device, and the first visual networking device can obtain the physical address of the first RDMA device based on the first key information, thereby encapsulating the physical address of the first RDMA device into the fifth data packet. In this way, the first visual networking device can send the fifth data packet to the first RDMA device based on the second data header information and according to the physical address of the first RDMA device.

[0107] Alternatively, as described in step D-2 above, the first visual networking device can obtain the IP address of the first RDMA device, and then the first visual networking device can obtain the physical address of the first RDMA device based on the IP address of the first RDMA device, thereby encapsulating the physical address of the first RDMA device into the fifth data packet. In this way, the first visual networking device can send the fifth data packet to the first RDMA device based on the second data header information and according to the physical address of the first RDMA device.

[0108] Optionally, in step E-1, obtaining the physical address of the first RDMA device according to the first key information includes:

[0109] According to the stored second correspondence between the physical address of the RDMA device and the key information, a physical address corresponding to the first key information of the first RDMA device is obtained and determined as the physical address of the first RDMA device.

[0110] From this, it can be seen that the first visual networking device can pre-record the second correspondence between the physical address and key information of each RDMA device connected to it. In this way, the first visual networking device can obtain the physical address of the first RDMA device based on the key information of the above-mentioned first RDMA device and the second correspondence.

[0111] Optionally, in step E-1, obtaining the physical address of the first RDMA device according to the IP address of the first RDMA device includes:

[0112] According to the seventh correspondence between the stored physical address and IP address of the RDMA device, a physical address corresponding to the IP address of the first RDMA device is obtained and determined as the physical address of the first RDMA device.

[0113] From this, it can be seen that the first visual networking device can pre-record the seventh correspondence between the physical address and IP address of each RDMA device connected to it. In this way, the first visual networking device can obtain the physical address of the first RDMA device based on the IP address of the above-mentioned first RDMA device and the seventh correspondence.

[0114] It should be noted that the physical address of the first RDMA device in the fifth data packet is the destination physical address; optionally, the fifth data packet also includes a source physical address, which is the physical address of the first visual network device.

[0115] Optionally, the first visual networking device can be connected to multiple RDMA devices. In this case, the first visual networking device can record the eighth correspondence between the physical address and port of each RDMA device. In this way, the first visual networking device can determine the corresponding port based on the physical address of the first RDMA device in the fifth data packet and the eighth correspondence, and then send the fifth data packet to the first RDMA device through the port.

[0116] In some embodiments of the present application, the method further includes the following steps F-1 to F-3:

[0117] Step F-1: receiving a sixth data packet sent by the second visual networking device, wherein the sixth data packet includes the physical address of the third visual networking device and third data content;

[0118] Step F-2: determining the first port information corresponding to the physical address of the third visual networking device based on the stored fourth correspondence between the physical address and the port information of the visual networking device;

[0119] Step F-3: Send the third data content to the third visual network device according to the first port information.

[0120] As can be seen from steps F-1 to F-3, after the first visual networking device receives the sixth data packet sent by the second visual networking device, it can determine the first port information based on the physical address of the third visual networking device in the sixth data packet and the aforementioned fourth correspondence, and then send the third data content to the third visual networking device based on the first port information. This shows that the first visual networking device can also realize data transmission between the second and third visual networking devices.

[0121] Optionally, the sixth data packet may also include third visual network protocol information, wherein the third visual network protocol information may include a source visual network number and a destination visual network number, the source visual network number is the visual network number of the second visual network device, and the destination visual network number is the visual network number of the third visual network device.

[0122] Optionally, sending the third data content to the third visual networking device according to the first port information includes:

[0123] According to the first port information, an eighth data packet is sent to the third visual network device, wherein the eighth data packet includes the third data content and the fourth visual network protocol information, and the fourth visual network protocol information may include a source visual network number and a destination visual network number, the source visual network number is the visual network number of the second visual network device, and the destination visual network number is the visual network number of the third visual network device.

[0124] Furthermore, the fourth visual networking protocol information can also contain verification information. In this way, after the third visual networking device receives the eighth data packet, it can obtain the corresponding IP address based on the visual networking number of the second visual networking device, and then transmit the third data content to the third largest model deployed on the third visual networking device based on the IP address and the verification information in the fourth visual networking protocol information.

[0125] In some embodiments of the present application, the method further includes the following steps H-1 to H-3:

[0126] Step H-1: receiving a seventh data packet sent by the first RDMA device, wherein the seventh data packet includes third data header information of the RDMA protocol and fourth data content, and the third data header information includes the IP address of the second RDMA device;

[0127] Step H-2: determining the second port information corresponding to the IP address of the second RDMA device according to the fifth correspondence between the stored IP address and port information of the RDMA device;

[0128] Step H-3: Send the fourth data content to the second RDMA device according to the second port information.

[0129] As can be seen from steps H-1 to H-3, after receiving the seventh data packet sent by the first RDMA device, the first visual networking device can determine the second port information based on the physical address of the second RDMA device in the seventh data packet and the fifth correspondence described above. Based on the second port information, the device can then send the fourth data content to the second RDMA device. This shows that the first visual networking device can also implement data transmission between different RDMA devices.

[0130] Optionally, the seventh data packet may further include a source IP address and a destination IP address, wherein the source IP address is the IP address of the first RDMA device, and the destination IP address is the IP address of the second RDMA device.

[0131] In the second aspect, the embodiment of the present application provides an improved intercommunication and fusion method based on VRB multi-protocol; applied to a second visual network device, such as Figure 2 As shown, the method may include the following steps 201 to 202:

[0132] Step 201: Receive a third data packet sent by a first visual networking device based on first visual networking protocol information.

[0133] The third data packet includes the first visual networking protocol information and first data content, and the first visual networking protocol information includes first key information for identifying the first RDMA device and second key information for identifying the second visual networking device.

[0134] In addition, after the first visual networking device receives the first data packet sent by the first RDMA device, it can obtain the first key information used to identify the first RDMA device and the second key information used to identify the second visual networking device, thereby removing the first data header information in the first data packet to obtain a second data packet containing the first data content, and then encapsulating the visual networking protocol for the second data packet, and processing the second data packet based on the first key information and the second key information to obtain the above-mentioned third data packet; and then based on the first visual networking protocol information, sending the third data packet to the second visual networking device, so that the second visual networking device transmits the first data content to the first large model deployed on the second visual networking device according to the first key information and the second key information.

[0135] Step 202: According to the first key information and the second key information, the first data content is transmitted to the first large model deployed on the second visual network device.

[0136] The first data content can be understood as data content that needs to be transmitted to the second visual networking device for processing by the first large model deployed on the second visual networking device; that is, it can be understood as the first data content including the input content of the first large model. It can be seen that in some embodiments of the present application, when the first data content of the first RDMA device needs to be processed by the first large model deployed on the second visual networking device, the first RDMA device can first transmit the first data content to the first visual networking device, and then the first visual networking device can transmit the first data content to the second visual networking device.

[0137] From the above, it can be seen that in some embodiments of the present application, visual networking devices and RDMA devices can be deployed in a large model cluster, and in the process of transmitting the first data packet from the first RDMA device to the second visual networking device, the first visual networking device can remove the original first data header information of the RDMA protocol in the first data packet, and encapsulate the visual networking protocol information including the first key information and the second key information for it, thereby obtaining the above-mentioned third data packet. After the second visual networking device receives the third data packet, it can transmit the first data content to the first large model for processing based on the first key information and the second key information in the third data packet.

[0138] As can be seen, in some embodiments of the present application, by encapsulating the visual networking protocol information including the first key information and the second key information in an RDMA protocol data packet, a first visual networking device can achieve data transmission between the first RDMA device and the second visual networking device in a large model cluster, thereby achieving the interoperability and integration of multiple protocols in the large model. Furthermore, by removing the original RDMA protocol first data header information in the first data packet, network transmission pressure can be reduced, thereby improving network transmission performance.

[0139] In some embodiments of the present application, the first visual networking protocol information further includes first verification information; and transmitting the first data content to the first large model deployed on the second visual networking device based on the first key information and the second key information includes the following steps J-1 to J-3:

[0140] Step J-1: obtaining an IP address corresponding to the first key information according to a first correspondence between the stored IP address of the RDMA device and the key information, and determining the IP address as the IP address of the first RDMA device;

[0141] Step J-2: According to the third correspondence between the stored IP address of the visual network device and the key information, obtain the IP address corresponding to the second key information and determine it as the IP address of the second visual network device

[0142] Step J-3: Transmit the first data content to the first large model based on the IP address of the first RDMA device, the IP address of the second visual networking device and the first verification information.

[0143] It can be seen from steps J-1 to J-3 that after the second visual networking device receives the above-mentioned third data packet, it can obtain the IP address of the first RDMA device based on the first key information in the third data packet and the above-mentioned first correspondence, and obtain the IP address of the second visual networking device based on the second key information in the third data packet and the above-mentioned third correspondence, and then transmit the first data content to the first large model based on the IP address of the first RDMA device, the IP address of the second visual networking device and the above-mentioned first verification information in the first visual networking protocol information.

[0144] In some embodiments of the present application, the method further includes the following steps L-1 to L-2:

[0145] Step L-1: Encapsulating the second data content with a visual networking protocol based on the first key information and the second key information to obtain a fourth data packet, wherein the fourth data packet includes the second visual networking protocol information and the second data content, and the second visual networking protocol information includes the first key information and the second key information;

[0146] Step L-2: Send the fourth data packet to the first visual networking device.

[0147] Among them, the second data content includes: the content output by the first large model after the first data content is input into the first large model; or, the second data content includes the data content actively sent by the second visual networking device to the first RDMA device (that is, the content that needs to be input into the second large model deployed on the first RDMA device).

[0148] It can be seen from steps L-1 to L-2 that the second data content needs to be transmitted from the second visual networking device to the first RDMA device. The second visual networking device can encapsulate the second visual networking protocol information for the second data content based on the first key information used to identify the first RDMA device and the second key information used to identify the second visual networking device, obtain the above-mentioned fourth data packet, and then send the fourth data packet to the first visual networking device based on the second visual networking protocol information, so that the first visual networking device can then send the second data content therein to the first RDMA device.

[0149] In addition, after the first visual networking device receives the fourth data packet, it can obtain the IP address corresponding to the first key information based on the first correspondence between the stored IP address of the RDMA device and the key information, and determine the IP address corresponding to the first key information as the IP address of the first RDMA device; obtain the IP address corresponding to the second key information based on the third correspondence between the stored IP address of the visual networking device and the key information, and determine the IP address corresponding to the second key information as the IP address of the second visual networking device; thereby processing the fourth data packet to obtain a fifth data packet, wherein the fifth data packet includes the second data header information of the RDMA protocol and the second data content, and the second data header information includes the IP address of the first RDMA device and the IP address of the second visual networking device; and then based on the second data header information, send the fifth data packet to the first RDMA device.

[0150] In some embodiments of the present application, the first visual networking device, the second visual networking device, and the third visual networking device are VRB devices.

[0151] In summary, the specific implementation of the VRB multi-protocol improved intercommunication and integration method according to the embodiment of the present application can be described as follows:

[0152] The improved intercommunication and fusion method based on VRB multi-protocol in this embodiment can be applied to Figure 3 The architecture shown in Figure 1. Figure 3 The architecture shown in FIG2 is shown in FIG3 , and the specific implementation process of the VRB multi-protocol improved intercommunication and integration method is as follows:

[0153] (1) The X-VRB device learns the MAC address, IP address, and port information of the connected RDMA device, and generates the KEY information of each RDMA device based on the IP address, thereby recording the MAC address, IP address, port information, and KEY information of the RDMA device connected to the X-VRB device into the dynamic learning table.

[0154] The X-VRB device learns the MAC address, IP address, and port information of the connected VRB device based on the connected VRB device, and generates the key information of each VRB device based on the IP address, thereby recording the MAC address, IP address, port information, and key information of the VRB device in the dynamic learning table.

[0155] The X-VRB device may synchronize the dynamic learning table to the VRB devices connected thereto (ie, the VRB1 device and the VRB2 device).

[0156] (2) The visual network protocol stack of the VRB1 device, the visual network protocol stack of the VRB2 device, and the visual network agent of the X-VRB device need to access the visual network. They can apply for access to the visual network based on the configured MAC address, thereby obtaining the visual network number. The following takes the visual network agent of the X-VRB device as an example to introduce the process of accessing the visual network:

[0157] The visual network server can send a broadcast data packet, which can carry at least one predetermined MAC, for example, MAC1. If the visual network agent needs to access the network through MAC1, it can encapsulate the visual network access application package according to MAC1 and send it to the visual network server. The visual network server determines the legitimacy of the visual network agent's MAC1 (that is, whether the MAC in the access application package is consistent with the locally stored MAC1). If it is legal (that is, the MAC in the access application package is consistent with the locally stored MAC1), it returns the visual network number and the sub-device number.

[0158] (3) The visual networking protocol stack of the VRB1 device, the visual networking protocol stack of the VRB2 device, and the visual networking agent of the X-VRB device negotiate the sub-device number of each device (i.e., the sub-device number used to establish the connection channel) through service transparent transmission. After the connection channel between the VRB1 device and the X-VRB device is established, the X-VRB device informs the VRB1 device of the device identification of the X-VRB device, and the VRB1 device saves the connection channel information (i.e., the sub-device number used) and the corresponding device identification of the X-VRB device to the visual networking protocol stack;

[0159] Similarly, after the connection channel between the VRB2 device and the X-VRB device is established, the X-VRB device informs the VRB2 device of the device identification of the X-VRB device, so that the VRB2 device saves the information of the connection channel (i.e. the sub-device number used) and the corresponding device identification of the X-VRB device to the visual network protocol stack.

[0160] It should be noted that: VRB devices and X-VRB devices are both visual networking devices. Each visual networking device has a visual networking number. One visual networking number corresponds to multiple sub-device numbers, that is, one visual networking device may have multiple sub-device numbers. To establish a channel between two visual networking devices, one sub-device number is required for each device. Therefore, when establishing a channel, it is necessary to negotiate which sub-device number to use to avoid the same sub-device number being occupied by multiple channels.

[0161] (4) Communication between RDMA1 device and VRB1 device:

[0162] When the first data content of the RDMA1 device needs to be transmitted to the VRB1 device and processed by the first large model deployed on the VRB1 device, the transmission process of the first data content is as follows: Figure 4 As shown:

[0163] The RDMA1 device encapsulates a first data packet based on the first data content, wherein the first data packet includes a source MAC address (i.e., the MAC address of the RDMA1 device) and a destination MAC address (i.e., the MAC address of the X-VRB device), an IP header, a UDP header, and the first data content, wherein the IP header includes a source IP address (i.e., the IP address of the RDMA1 device) and a destination IP address (i.e., the IP address of the VRB1 device);

[0164] The RDMA1 device sends a first data packet to the network card of the X-VRB device through its network card. The network card of the X-VRB device transmits the data packet to the visual networking agent of the X-VRB device. The visual networking agent transmits the data packet to the protocol identification unit. The protocol identification unit identifies that the first data packet is an RDMA protocol data packet and that the destination IP address is the IP address of the VRB1 device. The first data packet is then sent to the protocol conversion unit.

[0165] Furthermore, the protocol conversion unit extracts the IP address of the RDMA1 device in the first data packet, and searches the dynamic learning table stored in the dynamic learning unit for the KEY information corresponding to the IP address of the RDMA1 device. The IP address of the VRB1 device in the first data packet is extracted, and the KEY information corresponding to the IP address of the VRB1 device is searched in the dynamic learning table stored in the dynamic learning unit. The IP header and UDP header of the first data packet are removed, thereby encapsulating the visual networking protocol for the first data content based on the KEY information of the RDMA1 device and the KEY information of the VBR1 device to obtain a third data packet, and then transmitting the third data packet to the visual networking agent, wherein the third data packet includes: the first visual networking protocol information, the source MAC address (i.e., the MAC address of the X-VRB device) and the destination MAC address (i.e., the MAC address of the VRB1 device), and the first data content, wherein the first visual networking protocol information includes verification information and the KEY information of the RDMA1 device and the KEY information of the VRB1 device;

[0166] After receiving the third data packet, the visual network agent transmits the third data packet to the visual network protocol stack of the VRB1 device through the connection channel between the visual network agent and the visual network protocol stack of the VRB1 device (i.e., the channel formed by the visual network agent-network card of the X-VRB device-network card of the VRB1 device-visual network protocol stack);

[0167] After the visual networking protocol stack of the VRB1 device receives the third data packet, it extracts the KEY information of the RDMA1 device and the KEY information of the VRB1 device in the third data packet, and obtains the IP address corresponding to the KEY information of the RDMA1 device based on the stored dynamic learning table, and determines it as the IP address of the RDMA1 device, obtains the IP address corresponding to the KEY information of the VRB1 device, and determines it as the IP address of the VRB1 device, so that the kernel state driver of the VRB1 device encapsulates the IP header based on the IP address of the RDMA1 device and the IP address of the VRB1 device, and then based on the IP header and the verification information in the first visual networking protocol information, transmits the first data content to the first large model through the user state driver and the VRB application programming interface (API).

[0168] It should be noted that the kernel layer (i.e., the kernel-state driver) needs to use the IP address of the VBR1 device and the IP address of the RDMA1 device to convert between virtual memory and physical memory. In this embodiment, the third data packet includes the KEY information of the VRB1 device and the KEY information of the RDMA1 device. Therefore, after the kernel state receives the third data packet, it can search the dynamic learning table to obtain the IP address corresponding to the KEY information of the VRB1 device and the IP address corresponding to the KEY information of the RDMA1 device, so that the conversion between virtual memory and physical memory can be performed.

[0169] In addition, after receiving the third data packet, the kernel-mode driver needs to verify it (for example, verify its integrity), and can perform the verification based on the verification information in the first visual networking protocol information in the third data packet.

[0170] (5) Communication between VRB1 device and RDMA1 device:

[0171] After the first data content is input to the first large model deployed on the VRB1 device, the second data content is output. When the second data content needs to be returned to the RDMA1 device, the transmission process of the second data content is as follows: Figure 5 As shown:

[0172] After the first large model outputs the second data content, the second data content is transmitted to the kernel state driver of the VRB1 device, which is then transmitted to the visual networking protocol stack by the kernel state driver, so that the visual networking protocol stack generates second visual networking protocol information corresponding to the second data content, and the second visual networking protocol information includes the KEY information of the RDMA1 device and the KEY information of the VRB1 device, and then encapsulates the second visual networking protocol information and the second data content to obtain a fourth data packet, wherein the fourth data packet includes the second visual networking protocol information, the source MAC address (i.e., the MAC address of the VRB1 device) and the destination MAC address (i.e., the MAC address of the X-VRB device), and the second data content;

[0173] After receiving the third data packet, the visual networking protocol stack transmits the fourth data packet to the visual networking agent of the X-VRB device through the connection channel between the visual networking protocol stack and the visual networking agent of the X-VRB device;

[0174] The visual network agent transmits the fourth data packet to the protocol identification unit. The protocol identification unit identifies that the destination MAC address is the MAC address of the X-VRB device, and then transmits the fourth data packet to the protocol fine-tuning unit. The protocol fine-tuning unit extracts the KEY information of the VRB1 device and the KEY information of the RDMA1 device in the fourth data packet, and obtains the IP address corresponding to the KEY information of the VRB1 device and the MAC address and IP address corresponding to the KEY information of the RDMA1 device according to the stored dynamic learning table. An IP header and a UDP header are generated according to the IP address of the RDMA1 device and the IP address of the VRB1 device, and then the fifth data packet is encapsulated in combination with the MAC address of the RDMA1 device. The fifth data packet includes a source MAC address (i.e., the MAC address of the X-VRB device) and a destination MAC address (i.e., the MAC address of the RDMA1 device), an IP header, a UDP header, and a second data content. The IP header includes a source IP address (i.e., the IP address of the VRB1 device) and a destination IP address (i.e., the IP address of the RDMA1 device).

[0175] After the protocol fine-tuning unit obtains the fifth data packet, it returns it to the visual networking agent, which transmits it to the network card of the X-VRB device, and then transmits it to the network card of the RDMA1 device.

[0176] After the network card of the RDMA1 device receives the fifth data packet, it transmits the second data content in the fifth data packet to the second largest model deployed on the RDMA1 device through the TCP / IP protocol stack, kernel-mode driver, user-mode driver, and RDMA API in sequence.

[0177] It should be noted that when the VRB1 device actively sends data content to the RDMA1 device (for example, when the VRB1 device needs to send data content to the RDMA1 device so that the second largest model deployed on the RDMA1 device can be processed), the transmission process of the data content is the same as the transmission process of the second data content mentioned above, and will not be repeated here.

[0178] (6) Communication between VRB1 and VRB2:

[0179] When the third data content of the VRB 1 device needs to be transmitted to the VRB 2 device and processed by the third large model deployed on the VRB 2 device, the transmission process of the third data content is as follows: Figure 6 As shown:

[0180] The visual networking protocol stack of the VRB 1 device encapsulates a sixth data packet based on the third data content, wherein the sixth data packet includes the third visual networking protocol information, a source MAC address (i.e., the MAC address of the VRB 1 device), a destination MAC address (i.e., the MAC address of the X-VRB device), and the third data content, wherein the third visual networking protocol information includes a source visual networking number (i.e., the visual networking number of the VRB 1 device), a destination visual networking number (i.e., the visual networking number of the VRB 2 device), and verification information;

[0181] After receiving the sixth data packet, the visual networking protocol stack sends the sixth data packet to the visual networking agent of the X-VRB device through the connection channel between the visual networking protocol stack and the visual networking agent of the X-VRB device;

[0182] The visual networking agent transmits the sixth data packet to the protocol identification unit. The protocol identification unit identifies that the sixth data packet is a data packet of the visual networking protocol and that the destination device is the VRB2 device. The visual networking agent then searches the dynamic learning table for a port corresponding to the visual networking number of the VRB2 device, and based on the port, sends the eighth data packet to the visual networking protocol stack of the VRB2 device through the connection channel between the visual networking agent and the visual networking protocol stack of the VRB2 device. The eighth data packet includes: fourth visual networking protocol information, a source MAC address (i.e., the MAC address of the X-VRB device), a destination MAC address (i.e., the MAC address of the VRB2 device), and third data content. The fourth visual networking protocol information includes the source visual networking number (i.e., the visual networking number of the VRB1 device), the destination visual networking number (i.e., the visual networking number of the VRB2 device), and verification information.

[0183] Among them, after the visual networking protocol stack of the VRB2 device receives the eighth data packet, it extracts the visual networking number in the eighth data packet and converts it into an IP address. Based on the converted IP address and the verification information in the fourth visual networking protocol, the third data content is transmitted to the third largest model deployed on the VRB2 device through the kernel-state driver, user-state driver, and VRB API.

[0184] (7) Communication between RDMA1 device and RDMA2 device:

[0185] When the fourth data content of the RDMA1 device needs to be transmitted to the RDMA2 device and processed by the fourth large model deployed on the RDMA2 device, the transmission process of the fourth data content is as follows: Figure 7 As shown:

[0186] The RDMA1 device encapsulates a seventh data packet based on the fourth data content, where the seventh data packet includes a source MAC address (i.e., the MAC address of the RDMA1 device) and a destination MAC address (i.e., the MAC address of the X-VRB device), an IP header, a UDP header, and the fourth data content, where the IP header includes a source IP address (i.e., the IP address of the RDMA1 device) and a destination IP address (i.e., the IP address of the RDMA2 device).

[0187] The RDMA1 device sends the seventh data packet to the network card of the X-VRB device through its network card, which is then transmitted by the network card of the X-VRB device to the visual networking agent of the X-VRB device, and then transmitted by the visual networking agent to the protocol identification unit. The protocol identification unit identifies that the seventh data packet is a data packet of the RDMA protocol and that the destination IP address is the IP address of the RDMA2 device. The visual networking agent then searches the port corresponding to the IP address of the RDMA2 device in the dynamic learning table and transmits the ninth data packet to the network card of the X-VRB device. The ninth data packet is then sent to the network card of the RDMA2 device through the port corresponding to the IP address of the RDMA2 device on the network card of the VRB device. The ninth data packet includes a source MAC address (i.e., the MAC address of the X-VRB device) and a destination MAC address (i.e., the MAC address of the RDMA2 device), an IP header, a UDP header, and fourth data content. The IP header includes a source IP address (i.e., the IP address of the RDMA1 device) and a destination IP address (i.e., the IP address of the RDMA2 device).

[0188] After the network card of the RDMA1 device receives the ninth data packet, it transmits the fourth data content in the ninth data packet to the fourth model deployed on the RDMA2 device through the TCP / IP protocol stack, kernel-mode driver, user-mode driver, and RDMA API in sequence.

[0189] As can be seen from the above, in the embodiments of the present application, dynamic generation of KEY information is added during dynamic learning in the X-VRB device to facilitate IP information search. Furthermore, protocol identification is added. Since data transmission between multiple nodes in a large model uses a multi-connection method, protocol identification can quickly distinguish between identical protocols and then quickly process them. Furthermore, protocol fine-tuning is added. Specifically, VRB data does not contain UDP headers or IP header information. The protocol fine-tuning unit calculates the UDP header information based on the UDP algorithm and queries the IP information based on the KEY information. This allows the removal of IP information in the visual network, enabling multi-protocol data fusion and synchronization technology.

[0190] In summary (1) to (7) above, the VRB multi-protocol improved intercommunication and fusion method according to the embodiment of the present application is as follows: Figure 8 As shown, it mainly includes the following three aspects of management:

[0191] The first aspect: visual network management, which mainly deals with the legitimacy of devices in the visual network and the communication between devices; for example, the visual network protocol stack of VRB devices and the visual network agent of X-VRB devices need to enter the visual network to establish services such as video calls.

[0192] Among them, network entry and exit management can realize the legitimacy authentication of networked devices;

[0193] Keep-alive management is a dynamic detection of visual networking devices;

[0194] Data management is the management of data from the data transmission level, that is, data management based on different data requirements;

[0195] Business management is the management of various businesses to ensure the normal operation of various businesses, such as meetings, video calls, live broadcasts, etc.

[0196] The second aspect: X-VRB management, which mainly handles the conversion between VRB protocol and RoCE v2 protocol, including protocol identification management, protocol conversion management, protocol fine-tuning management, visual network agent, and dynamic learning management;

[0197] Among them, protocol identification management mainly identifies VRB protocol and RoCE v2 protocol;

[0198] Protocol conversion management is responsible for the conversion between RoCE v2 protocol and VRB protocol;

[0199] Protocol fine-tuning management is responsible for supplementing the required UDP header and IP header when the VRB protocol converts the RoCE v2 protocol;

[0200] The visual network agent has the visual network management function. In order to optimize the configuration of network resources, the visual network management layer data (such as network access, heartbeat, control messages, etc.) is managed through the management port (i.e. the corresponding port on the network card of the X-VRB device);

[0201] Dynamic IP learning management refers to: quickly matching the MAC address, IP address, and port number of the access device (i.e., the RDMA device and VRB device connected to X-VRB), generating key information based on the IP address, and storing the matching results and corresponding data in the dynamic learning table, so as to achieve fast data forwarding from VRB to RoCE v2 data.

[0202] The third aspect: VRB management, including network connection management, learning table management, event synchronization management, network quality management, and log management;

[0203] Network connection management refers to: visual network management;

[0204] Learning table management means that during data interaction with an X-VRB device, the required IP address can be obtained based on the dynamic learning table. This allows data to be encapsulated and sent based on the IP address, enabling data transmission between VRB devices and X-VRB devices.

[0205] Event synchronization management, responsible for synchronizing data to the upper layer when receiving or sending data;

[0206] Network quality management, responsible for the detection, regulation and load balancing of network quality during VRB device transmission;

[0207] Log management records the operation status of VRB equipment.

[0208] In summary, in the VRB-based multi-protocol improved intercommunication and fusion method of the present embodiment, the X-VRB device is responsible for establishing a connection and data communication with the VRB device side, and for communicating with the RDMA device side. It converts received RDMA RoCE v2 protocol frames into VRB protocol frame data and sends it to the VRB device. It also converts received VRB data into RoCEv2 protocol frames and sends them to the RDMA device, thereby achieving data fusion and synchronization.

[0209] It's also worth noting that the current visual networking environment uses VRB to implement RDMA technology, using X-VRB devices to access the RoCE v2 protocol for RDMA, enabling interoperability and integration between visual networking and Internet RDMA technology. However, this technology still uses IP data structures during visual networking transmission. The presence of IP structures not only occupies multiple bytes during transmission, but also renders IP data transmission structures meaningless due to non-IP addressing in visual networking. This also increases network transmission redundancy, increasing network transmission pressure and degrading network transmission performance.

[0210] The embodiments of the present application improve the dynamic learning function and protocol fine-tuning function of the X-VRB device, remove IP and other information from the RoCE v2 data conversion, complete IP-free data transmission in the visual network environment, and realize the intercommunication, integration and synchronization of protocols such as VRB and RoCE v2, thereby reducing network transmission pressure and improving network transmission performance.

[0211] In addition, the aforementioned Remote Direct Memory Access (RDMA) technology is a new memory access technology that allows servers to directly read and write memory data of other servers at high speed without the need for time-consuming processing by the operating system / CPU.

[0212] The Ethernet-based Remote Direct Memory Access (RDMA) protocol (RDMA over Converged Ethernet version 2, RoCE v2) is designed to achieve high-performance, low-latency data transmission over an Ethernet network. RoCE v2 integrates Ethernet and RDMA technologies, significantly enhancing performance and flexibility, and is particularly suitable for application scenarios that require extremely high data interaction speed and efficiency. It should be noted that for the method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited to the order of the actions described, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0213] In a third aspect, the embodiment of the present application provides an improved intercommunication and fusion device based on VRB multi-protocol; applied to a first visual network device, such as Figure 9 As shown, the device may include the following modules:

[0214] A first receiving module 901 is configured to receive a first data packet sent by a first RDMA device, wherein the first data packet includes first data header information and first data content of an RDMA protocol, the first data header information includes an IP address of a second visual network device, and RDMA stands for Remote Direct Memory Access;

[0215] A first acquiring module 902 is configured to acquire first key information for identifying the first RDMA device;

[0216] A second acquisition module 903 is configured to acquire second key information for identifying the second visual network device;

[0217] a removal module 904, configured to remove the first data header information from the first data packet to obtain a second data packet;

[0218] A first processing module 905 is configured to encapsulate a visual networking protocol for the second data packet, and process the second data packet based on the first key information and the second key information to obtain a third data packet, wherein the third data packet includes the first visual networking protocol information and the first data content, and the first visual networking protocol information includes the first key information and the second key information;

[0219] The first sending module 906 is used to send the third data packet to the second visual networking device based on the first visual networking protocol information, so that the second visual networking device transmits the first data content to the first large model deployed on the second visual networking device according to the first key information and the second key information.

[0220] Optionally, the first obtaining module 902 is specifically configured to:

[0221] When the first data header information further includes the IP address of the first RDMA device, obtaining the first key information corresponding to the IP address of the first RDMA device according to a stored first correspondence between the IP address of the RDMA device and the key information;

[0222] or,

[0223] In a case where the first data packet further includes the physical address of the first RDMA device, the first key information corresponding to the physical address of the first RDMA device is acquired according to a stored second correspondence between the physical address of the RDMA device and the key information.

[0224] Optionally, the second obtaining module 903 is specifically configured to:

[0225] According to the third correspondence between the stored IP address of the visual networking device and the key information, the second key information corresponding to the IP address of the second visual networking device is obtained.

[0226] Optionally, the device further comprises:

[0227] A third acquisition module, configured to acquire a physical address of the second visual networking device according to the IP address of the second visual networking device;

[0228] A first encapsulation module, configured to encapsulate the physical address of the second visual networking device into the third data packet;

[0229] The first sending module 906 is specifically configured to:

[0230] Based on the first visual networking protocol information and the physical address of the second visual networking device, the third data packet is sent to the second visual networking device.

[0231] Optionally, the device further comprises:

[0232] a third receiving module, configured to receive a fourth data packet sent by the second visual networking device based on the second visual networking protocol information, wherein the fourth data packet includes the second visual networking protocol information and second data content, and the second visual networking protocol information includes the first key information and the second key information;

[0233] a fourth acquisition module, configured to acquire an IP address corresponding to the first key information according to a first correspondence between the stored IP address of the RDMA device and the key information, and determine the IP address corresponding to the first key information as the IP address of the first RDMA device;

[0234] a fifth acquisition module, configured to acquire the IP address corresponding to the second key information based on a third correspondence between the stored IP address of the visual network device and the key information, and determine the IP address corresponding to the second key information as the IP address of the second visual network device;

[0235] a second processing module, configured to process the fourth data packet to obtain a fifth data packet, wherein the fifth data packet includes second data header information of the RDMA protocol and the second data content, and the second data header information includes the IP address of the first RDMA device and the IP address of the second visual networking device;

[0236] The third sending module is configured to send the fifth data packet to the first RDMA device based on the second data header information.

[0237] Optionally, the device further comprises:

[0238] a sixth acquiring module, configured to acquire a physical address of the first RDMA device according to the first key information or the IP address of the first RDMA device;

[0239] A second encapsulation module, configured to encapsulate the physical address of the first RDMA device into the fifth data packet;

[0240] The third sending module is specifically configured to:

[0241] The fifth data packet is sent to the first RDMA device based on the second data header information and the physical address of the first RDMA device.

[0242] Optionally, the device further comprises:

[0243] a fourth receiving module, configured to receive a sixth data packet sent by the second visual networking device, wherein the sixth data packet includes a physical address of the third visual networking device and third data content;

[0244] A first determining module, configured to determine the first port information corresponding to the physical address of the third visual networking device according to a fourth correspondence between the stored physical address and port information of the visual networking device;

[0245] The fourth sending module is used to send the third data content packet to the third visual network device according to the first port information.

[0246] Optionally, the device further comprises:

[0247] a fifth receiving module, configured to receive a seventh data packet sent by the first RDMA device, wherein the seventh data packet includes third data header information of the RDMA protocol and fourth data content, and the third data header information includes the IP address of the second RDMA device;

[0248] A second determining module is configured to determine the second port information corresponding to the IP address of the second RDMA device according to a fifth correspondence between the stored IP address and port information of the RDMA device;

[0249] A fifth sending module is configured to send the fourth data content to the second RDMA device according to the second port information.

[0250] In a fourth aspect, the embodiment of the present application provides an improved intercommunication and fusion device based on VRB multi-protocol; Figure 10 As shown, the device may include the following modules:

[0251] A second receiving module 1001 is configured to receive a third data packet sent by the first visual networking device based on the first visual networking protocol information, wherein the third data packet includes the first visual networking protocol information and the first data content, and the first visual networking protocol information includes first key information for identifying the first RDMA device and second key information for identifying the second visual networking device;

[0252] The second sending module 1002 is used to transmit the first data content to the first large model deployed on the second visual network device according to the first key information and the second key information.

[0253] Optionally, the first visual networking protocol information further includes first verification information; the second sending module 1002 is specifically configured to:

[0254] According to a first correspondence between the stored IP address of the RDMA device and the key information, obtaining the IP address corresponding to the first key information and determining it as the IP address of the first RDMA device;

[0255] According to the third correspondence between the stored IP address of the visual network device and the key information, the IP address corresponding to the second key information is obtained and determined as the IP address of the second visual network device.

[0256] The first data content is transmitted to the first large model according to the IP address of the first RDMA device, the IP address of the second visual networking device and the first verification information.

[0257] Optionally, the device further comprises:

[0258] a third encapsulation module, configured to encapsulate a visual networking protocol for the second data content based on the first key information and the second key information to obtain a fourth data packet, wherein the fourth data packet includes the second visual networking protocol information and the second data content, and the second visual networking protocol information includes the first key information and the second key information;

[0259] The sixth sending module is used to send the fourth data packet to the first visual network device.

[0260] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0261] In the fifth aspect, an embodiment of the present application also provides an improved intercommunication and fusion system based on VRB multi-protocol, including a first visual networking device and a second visual networking device, wherein the first visual networking device executes the improved intercommunication and fusion method based on VRB multi-protocol described in the first aspect above, and the second visual networking device executes the improved intercommunication and fusion method based on VRB multi-protocol described in the second aspect above.

[0262] In a sixth aspect, an embodiment of the present application further provides an electronic device, including:

[0263] one or more processors; and

[0264] One or more machine-readable media having instructions stored thereon, when executed by the one or more processors, cause the electronic device to execute the improved intercommunication fusion method based on VRB multi-protocol described in the first aspect above, or execute the improved intercommunication fusion method based on VRB multi-protocol described in the second aspect above.

[0265] In the seventh aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program that enables a processor to execute the improved intercommunication and fusion method based on VRB multi-protocol described in the first aspect above, or to execute the improved intercommunication and fusion method based on VRB multi-protocol described in the second aspect above.

[0266] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0267] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, devices, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0268] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0269] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0270] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0271] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0272] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or terminal device that includes the element.

[0273] The above is a detailed introduction to the VRB multi-protocol improved intercommunication and fusion method, device, electronic device, and storage medium provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only intended to help understand the method and core ideas of the present application. At the same time, for those skilled in the art, according to the ideas of the present application, there may be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as limiting the present application.

Claims

1. An improved VRB multi-protocol intercommunication and integration method, characterized in that: Applied to a first visual networking device, the method includes: Receive a first data packet sent by a first RDMA device, wherein the first data packet includes first data header information and first data content of an RDMA protocol, the first data header information includes an IP address of a second visual network device, and RDMA stands for remote direct memory access; Acquire first key information for identifying the first RDMA device; Acquire second key information for identifying the second visual network device; removing the first data header information from the first data packet to obtain a second data packet; encapsulating a visual networking protocol for the second data packet, and processing the second data packet based on the first key information and the second key information to obtain a third data packet, wherein the third data packet includes the first visual networking protocol information and the first data content, and the first visual networking protocol information includes the first key information and the second key information; Based on the first visual networking protocol information, the third data packet is sent to the second visual networking device, so that the second visual networking device transmits the first data content to the first large model deployed on the second visual networking device according to the first key information and the second key information.

2. The method according to claim 1, characterized in that The acquiring first key information for identifying the first RDMA device includes: When the first data header information further includes the IP address of the first RDMA device, obtaining the first key information corresponding to the IP address of the first RDMA device according to a stored first correspondence between the IP address of the RDMA device and the key information; or, In a case where the first data packet further includes the physical address of the first RDMA device, the first key information corresponding to the physical address of the first RDMA device is acquired according to a stored second correspondence between the physical address of the RDMA device and the key information.

3. The method according to claim 1, characterized in that The acquiring of second key information for identifying the second visual network device includes: According to the third correspondence between the stored IP address of the visual networking device and the key information, the second key information corresponding to the IP address of the second visual networking device is obtained.

4. The method according to any one of claims 1 to 3, characterized in that The method further comprises: Obtaining a physical address of the second visual networking device according to the IP address of the second visual networking device; Encapsulating the physical address of the second visual networking device into the third data packet; The sending the third data packet to the second visual networking device based on the first visual networking protocol information includes: Based on the first visual networking protocol information and the physical address of the second visual networking device, the third data packet is sent to the second visual networking device.

5. The method according to any one of claims 1 to 3, characterized in that The method further comprises: receiving a fourth data packet sent by the second visual networking device based on the second visual networking protocol information, wherein the fourth data packet includes the second visual networking protocol information and second data content, and the second visual networking protocol information includes the first key information and the second key information; Acquire the IP address corresponding to the first key information according to a first correspondence between the stored IP address of the RDMA device and the key information, and determine the IP address corresponding to the first key information as the IP address of the first RDMA device; Obtaining the IP address corresponding to the second key information according to the third correspondence between the stored IP address of the visual network device and the key information, and determining the IP address corresponding to the second key information as the IP address of the second visual network device; Processing the fourth data packet to obtain a fifth data packet, wherein the fifth data packet includes second data header information of the RDMA protocol and the second data content, and the second data header information includes the IP address of the first RDMA device and the IP address of the second visual networking device; Based on the second data header information, the fifth data packet is sent to the first RDMA device.

6. The method according to claim 5, characterized in that The method further comprises: Acquire a physical address of the first RDMA device according to the first key information or the IP address of the first RDMA device; Encapsulating the physical address of the first RDMA device into the fifth data packet; The sending the fifth data packet to the first RDMA device based on the second data header information includes: The fifth data packet is sent to the first RDMA device based on the second data header information and the physical address of the first RDMA device.

7. The method according to any one of claims 1 to 3, characterized in that The method further comprises: receiving a sixth data packet sent by the second visual networking device, wherein the sixth data packet includes a physical address of the third visual networking device and third data content; Determining the first port information corresponding to the physical address of the third visual networking device according to the stored fourth correspondence between the physical address and the port information of the visual networking device; According to the first port information, the third data content packet is sent to the third visual network device.

8. The method according to any one of claims 1 to 3, characterized in that The method further comprises: receiving a seventh data packet sent by the first RDMA device, wherein the seventh data packet includes third data header information of the RDMA protocol and fourth data content, and the third data header information includes the IP address of the second RDMA device; Determining the second port information corresponding to the IP address of the second RDMA device according to the fifth correspondence between the stored IP address and the port information of the RDMA device; The fourth data content is sent to the second RDMA device according to the second port information.

9. An improved intercommunication and integration method based on VRB multi-protocol, characterized in that: Applied to a second visual networking device, the method includes: Receiving a third data packet sent by the first visual networking device based on the first visual networking protocol information, wherein the third data packet includes the first visual networking protocol information and the first data content, and the first visual networking protocol information includes first key information for identifying the first RDMA device and second key information for identifying the second visual networking device; According to the first key information and the second key information, the first data content is transmitted to the first large model deployed on the second visual network device.

10. The method according to claim 9, characterized in that The first visual networking protocol information further includes first verification information; and transmitting the first data content to the first large model deployed on the second visual networking device according to the first key information and the second key information includes: According to a first correspondence between the stored IP address of the RDMA device and the key information, obtaining the IP address corresponding to the first key information and determining it as the IP address of the first RDMA device; According to the third correspondence between the stored IP address of the visual network device and the key information, the IP address corresponding to the second key information is obtained and determined as the IP address of the second visual network device. The first data content is transmitted to the first large model according to the IP address of the first RDMA device, the IP address of the second visual networking device and the first verification information.

11. The method according to claim 9 or 10, characterized in that The method further comprises: Encapsulating a visual networking protocol for the second data content based on the first key information and the second key information to obtain a fourth data packet, wherein the fourth data packet includes the second visual networking protocol information and the second data content, and the second visual networking protocol information includes the first key information and the second key information; Send the fourth data packet to the first visual networking device.

12. An improved intercommunication and fusion device based on VRB multi-protocol, characterized in that: Applied to a first visual networking device, the apparatus includes: A first receiving module is configured to receive a first data packet sent by a first RDMA device, wherein the first data packet includes first data header information and first data content of an RDMA protocol, the first data header information includes an IP address of a second visual network device, and RDMA stands for Remote Direct Memory Access; A first acquisition module, configured to acquire first key information for identifying the first RDMA device; A second acquisition module, configured to acquire second key information for identifying the second visual network device; a removal module, configured to remove the first data header information from the first data packet to obtain a second data packet; a first processing module, configured to encapsulate a visual networking protocol for the second data packet, and process the second data packet based on the first key information and the second key information to obtain a third data packet, wherein the third data packet includes the first visual networking protocol information and the first data content, and the first visual networking protocol information includes the first key information and the second key information; The first sending module is used to send the third data packet to the second visual networking device based on the first visual networking protocol information, so that the second visual networking device transmits the first data content to the first large model deployed on the second visual networking device according to the first key information and the second key information.

13. An improved intercommunication and fusion device based on VRB multi-protocol, characterized in that: Applied to a second visual networking device, the apparatus comprises: a second receiving module, configured to receive a third data packet sent by the first visual networking device based on the first visual networking protocol information, wherein the third data packet includes the first visual networking protocol information and the first data content, and the first visual networking protocol information includes first key information for identifying the first RDMA device and second key information for identifying the second visual networking device; The second sending module is used to transmit the first data content to the first large model deployed on the second visual network device according to the first key information and the second key information.

14. An electronic device, characterized in that: include: one or more processors; and One or more machine-readable media having instructions stored thereon, when executed by the one or more processors, cause the electronic device to execute the VRB multi-protocol improved intercommunication fusion method based on any one of claims 1 to 8, or execute the VRB multi-protocol improved intercommunication fusion method based on any one of claims 9 to 11.

15. A computer-readable storage medium, characterized in that The computer program stored therein enables the processor to execute the improved intercommunication fusion method based on VRB multi-protocol according to any one of claims 1 to 8, or to execute the improved intercommunication fusion method based on VRB multi-protocol according to any one of claims 9 to 11.

Images