WO2024149043A1 - Reliable transmission method and apparatus for p2mp data - Google Patents

Abstract

A reliable transmission method for P2MP data, which method is applied to a first node, wherein the first node is a node in a P2MP communication domain, the P2MP communication domain comprises a plurality of nodes, the first node is one of the plurality of nodes, and a logic interconnection network is established between the plurality of nodes. The method comprises: receiving a P2MP data packet sent by a second node in a P2MP communication domain, wherein the P2MP data packet at least comprises P2MP data, and the second node is one of a plurality of nodes; determining a neighbor node in a logic interconnection network according to a forwarding table of a first node; and sending a response message to the neighbor node according to the state of the receiving of the P2MP data, and the neighbor node executing a reliable transmission processing task with regard to the first node receiving the P2MP data. In this way, a logic interconnection network is established between a plurality of nodes in a P2MP communication domain, and the task, which is executed by only a source node, of reliable transmission processing of P2MP data is deployed on the plurality of nodes in the P2MP communication domain in a distributed manner, such that the single-point processing pressure can be effectively reduced.

Description

A reliable transmission method and device for P2MP data

This application claims priority to a Chinese patent application filed with the State Intellectual Property Office of China on January 10, 2023, with application number 202310037879.6 and application name “A method and device for reliable transmission of P2MP data”, the entire contents of which are incorporated by reference into this application.

Technical Field

The present application relates to the field of communication network technology, and in particular to a reliable transmission method and device for P2MP data.

Background technique

Point-to-multipoint (P2MP) communication refers to the communication between a source node that sends multicast data and a series of receiving nodes that receive the multicast data.

At present, the reliable guarantee mechanism for P2MP data transmission/reception usually establishes a unicast reliable transmission connection at the transport layer between the source node and each receiving node of the P2MP data, and guarantees reliable transmission through the reliable transmission connection, or establishes a reliable transmission mechanism at the application layer based on the unreliable datagram connection at the transport layer to ensure reliable transmission. However, when the scale of communication nodes in the P2MP communication domain is large, the resource consumption required for the source node to maintain the reliable transmission mechanism alone may be serious; or when the multicast source frequently switches in the P2MP communication domain, the overhead of establishing a reliable transmission connection for P2MP data on the control plane may be too high, which ultimately leads to the inability to guarantee the reliability of multicast data transmission. Therefore, it is necessary to establish an efficient P2MP data reliable transmission guarantee mechanism.

Summary of the invention

The present application provides a reliable transmission method, device, electronic device, computer-readable storage medium and computer program product for P2MP data, which can efficiently ensure the reliable transmission of P2MP data when the scale of communication nodes in the P2MP communication domain is large and multicast source switching occurs frequently in the P2MP communication domain.

In a first aspect, the present application provides a reliable transmission method for P2MP data, which is applied to a first node, wherein the first node is a node in a P2MP communication domain, wherein the P2MP communication domain includes multiple nodes, the first node is one of the multiple nodes, a logical interconnection network is established between the multiple nodes, and each node has its own identification information. The method includes: receiving a P2MP data packet sent by a second node in a P2MP communication domain; the P2MP data packet includes at least P2MP data, and the second node is one of the multiple nodes; determining the identification information of a third node according to a forwarding table of the first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node having a direct connection relationship in the logical interconnection network; and sending a response message to the third node according to a state of receiving the P2MP data and the identification information of the third node.

Therefore, by establishing a logical interconnection network for multiple nodes in the P2MP communication domain, the task of reliable transmission and processing of P2MP data performed only by the source node is distributed and deployed to multiple nodes in the P2MP communication domain, which can effectively reduce the pressure of single-point processing.

In a possible implementation, the P2MP data packet includes at least the P2MP data and the sequence number of the P2MP data; sending a response message to the third node according to the state of receiving the P2MP data and the identification information of the third node includes: when the state of receiving the P2MP data is erroneous, sending the response message to the third node; the response message includes at least the identification information of the first node, the sequence number of the P2MP data and a state flag of receiving the P2MP data, the state flag of receiving the P2MP data is set to NAK, and NAK indicates that the received P2MP data is erroneous; the method also includes: receiving a correct P2MP data packet retransmitted by the third node.

In a possible implementation, the P2MP data packet includes at least P2MP data and identification information of the second node; determining the identification information of the third node according to the forwarding table of the first node includes: determining the identification information of the third node according to the forwarding table of the first node and the identification information of the second node; wherein the forwarding table of the first node contains at least one field {key: value}, the key in the field is the identification information of the second node or a wildcard, and the value in the field is the identification information of the third node.

In one possible implementation, a logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including: each of the multiple nodes has at least one neighbor node, each of the multiple nodes establishes a direct connection with the neighbor node, and no direct connection is established between the neighbor nodes of each of the multiple nodes.

In a possible implementation, the logical interconnection network is established by connecting multiple nodes. It includes: a plurality of nodes are divided into a plurality of node groups, each node group includes a first node, the physical distance between the nodes in the node group is less than or equal to a preset distance, connections are established between the nodes in the node group, and the first nodes in the plurality of node groups are connected.

In a second aspect, the present application provides a reliable transmission method for P2MP data, which is applied to a third node, the third node is a node in a P2MP communication domain, a P2MP communication domain includes multiple nodes, the third node is one of the multiple nodes, a logical interconnection network is established between the multiple nodes, and each node has its own identification information. The method includes: receiving a response message from a first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node with a direct connection relationship in the logical interconnection network; the response message includes at least the identification information of the first node and a status flag of the first node receiving P2MP data; the P2MP data is included in a P2MP data packet, and the P2MP data packet is sent by a second node in a P2MP communication domain, and the second node is one of the multiple nodes; according to the response message of the first node, the management table of the third node is updated; the management table includes at least the status of the first node receiving P2MP data.

Therefore, by establishing a logical interconnection network for multiple nodes in the P2MP communication domain, the task of reliable transmission and processing of P2MP data performed only by the source node is distributed and deployed to multiple nodes in the P2MP communication domain, which can effectively reduce the pressure of single-point processing.

In a possible implementation, after receiving a response message from the first node, the method further includes: determining whether a status flag of the first node receiving P2MP data in the response message of the first node is NAK, and if so, retransmitting a correct P2MP data packet to the first node and starting a timer; wherein NAK indicates that the P2MP data received by the first node is incorrect; receiving a response message from the first node, and stopping the timer.

In a possible implementation, the P2MP data packet includes at least P2MP data and identification information of the second node; updating the management table of the third node according to the response message of the first node includes: updating the state flag of the first node receiving the P2MP data in the management table of the third node according to the identification information of the first node and the state flag of the first node receiving the P2MP data; wherein the management table of the second node includes at least one field {key: value: state}, the key in the field is the identification information or a wildcard of the second node, the value in the field is the identification information of the first node, and the state in the field is the state flag of the first node receiving the P2MP data.

In one possible implementation, a logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including: each of the multiple nodes has at least one neighbor node, each of the multiple nodes establishes a direct connection with the neighbor node, and no direct connection is established between the neighbor nodes of each of the multiple nodes.

In one possible implementation, a logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including: multiple nodes are divided into multiple node groups, each node group includes a third node, the physical distance between the nodes in the node group is less than or equal to a preset distance, the nodes in the node group are connected, and the third nodes in multiple node groups are connected.

In a third aspect, the present application provides a reliable transmission device for P2MP data, which is deployed on a first node, the first node is a node in a P2MP communication domain, a P2MP communication domain includes multiple nodes, the first node is one of the multiple nodes, a logical interconnection network is established between the multiple nodes, and each node has its own identification information. The device includes: a communication module, which is used to receive a P2MP data packet sent by a second node in a P2MP communication domain; the P2MP data packet includes at least P2MP data, and the second node is one of the multiple nodes; a processing module, which is used to determine the identification information of a third node according to a forwarding table of the first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node with a direct connection relationship in the logical interconnection network; the processing module is also used to send a response message to the third node according to the state of receiving the P2MP data and the identification information of the third node.

In a possible implementation, the P2MP data packet includes at least the P2MP data and the sequence number of the P2MP data; when the processing module sends a response message to the third node according to the state of receiving the P2MP data and the identification information of the third node, it is used to: when the state of receiving the P2MP data is erroneous, send a response message to the third node; the response message includes at least the identification information of the first node, the sequence number of the P2MP data and a state flag of receiving the P2MP data, and the state flag of receiving the P2MP data is set to NAK, and NAK indicates that the received P2MP data is erroneous; the communication module is also used to: receive a correct P2MP data packet retransmitted by the third node.

In a possible implementation, the P2MP data packet includes at least P2MP data and identification information of the second node; when the processing module determines the identification information of the third node according to the forwarding table of the first node, it is used to: determine the identification information of the third node according to the forwarding table of the first node and the identification information of the second node; wherein the forwarding table of the first node includes at least one field {key: value}, the key in the field is the identification information of the second node or a wildcard, and the value in the field is the identification information of the third node.

In one possible implementation, a logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including: each of the multiple nodes has at least one neighbor node, each of the multiple nodes establishes a direct connection with the neighbor node, and no direct connection is established between the neighbor nodes of each of the multiple nodes.

In one possible implementation, a logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including: multiple nodes are divided into multiple node groups, each node group includes a first node, the physical distance between the nodes in the node group is less than or equal to a preset distance, connections are established between the nodes in the node group, and connections are established with the first nodes in the multiple node groups.

In a fourth aspect, the present application provides a reliable transmission device for P2MP data, which is deployed on a third node, the third node is a node in a P2MP communication domain, a P2MP communication domain includes multiple nodes, the third node is one of the multiple nodes, a logical interconnection network is established between the multiple nodes, and each node has its own identification information. The device includes: a communication module, which is used to receive a response message from a first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node with a direct connection relationship in the logical interconnection network; the response message includes at least the identification information of the first node and a status flag of the first node receiving P2MP data; the P2MP data is included in a P2MP data packet, and the P2MP data packet is sent by the second node in a P2MP communication domain, and the second node is one of the multiple nodes; a processing module, which is used to update the management table of the third node according to the response message of the first node; the management table includes at least the status of the first node receiving P2MP data.

In a possible implementation, after receiving the response message from the first node, the processing module is further used to: determine whether the status flag of the first node receiving the P2MP data in the response message of the first node is NAK, and if so, retransmit the correct P2MP data packet to the first node and start the timer; wherein NAK indicates that the P2MP data received by the first node is incorrect; and the communication module is further used to: receive the response message from the first node and stop the timer.

In a possible implementation, the P2MP data packet includes at least P2MP data and identification information of the second node; when the processing module updates the management table of the second node according to the response message of the first node, it is used to: update the state flag of the first node receiving the P2MP data in the management table of the second node according to the identification information of the first node and the state flag of the first node receiving the P2MP data; wherein the management table of the second node includes at least one field {key: value: state}, the key in the field is the identification information or a wildcard of the second node, the value in the field is the identification information of the first node, and the state in the field is the state flag of the first node receiving the P2MP data.

In one possible implementation, a logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including: each of the multiple nodes has at least one neighbor node, each of the multiple nodes establishes a direct connection with the neighbor node, and no direct connection is established between the neighbor nodes of each of the multiple nodes.

In one possible implementation, a logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including: multiple nodes are divided into multiple node groups, each node group includes a third node, the physical distance between the nodes in the node group is less than or equal to a preset distance, the nodes in the node group are connected, and the third nodes in multiple node groups are connected.

In a fifth aspect, the present application provides an electronic device, comprising: at least one memory for storing programs; and at least one processor for executing the programs stored in the memory; wherein, when the program stored in the memory is executed, the processor is used to execute the method described in the first aspect or any possible implementation of the first aspect, or to execute the method described in the second aspect or any possible implementation of the second aspect.

In a sixth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect, or executes the method described in the second aspect or any possible implementation of the second aspect.

In the seventh aspect, the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method described in the first aspect or any possible implementation of the first aspect, or executes the method described in the second aspect or any possible implementation of the second aspect.

It can be understood that the beneficial effects of the third to seventh aspects mentioned above can be found in the relevant descriptions of the first to second aspects mentioned above, and will not be repeated here.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of an encoding process for establishing an MPI communication domain;

FIG2 is a schematic diagram of a typical MPI interface using a P2MP communication model;

FIG3 is a flow chart of network layer multicast forwarding and application layer multicast forwarding P2MP data;

FIG4 is a schematic diagram of the salient features of P2MP data forwarding within a P2MP communication domain;

FIG5 is a schematic diagram illustrating the problem of reliable forwarding of P2MP data;

FIG6a is a physical interconnection network established between nodes in a P2MP communication domain provided by an embodiment of the present application;

FIG6b is a forwarding path of P2MP data in a physical interconnection network provided by an embodiment of the present application;

FIG7a is a logical interconnection network established between nodes in a P2MP communication domain provided by an embodiment of the present application;

FIG7b is a reliable transmission confirmation relationship of P2MP data in a logical interconnection network provided by an embodiment of the present application;

FIG8 is a logical interconnection network established based on a quick response strategy provided in an embodiment of the present application;

FIG9 is a flow chart of a reliable transmission method for P2MP data provided in an embodiment of the present application;

FIG10 is a flow chart of a reliable transmission method for P2MP data provided in an embodiment of the present application;

FIG11 is a schematic diagram of a logical interconnection network and a forwarding table provided in an embodiment of the present application;

FIG12 is a flow chart of a reliable transmission method for P2MP data provided in an embodiment of the present application;

FIG13a is a diagram of a network layer multicast resource configuration based on RC connection in a P2MP communication domain provided by an embodiment of the present application;

FIG13b is a schematic diagram of a logical connection based on a data plane in a P2MP communication domain provided by an embodiment of the present application;

FIG13c is a schematic diagram of a logical connection based on a control plane in a P2MP communication domain provided by an embodiment of the present application;

FIG14 is a schematic diagram of a reliable transmission device for P2MP data provided in an embodiment of the present application;

FIG. 15 is a schematic diagram of a reliable transmission device for P2MP data provided in an embodiment of the present application.

Detailed ways

The term "and/or" in this article is a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and/or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The symbol "/" in this article indicates that the associated objects are in an or relationship, for example, A/B means A or B.

The terms "first" and "second" in the specification and claims herein are used to distinguish different objects rather than to describe a specific order of the objects. For example, a first response message and a second response message are used to distinguish different response messages rather than to describe a specific order of the response messages.

In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more than two. For example, multiple processing units refer to two or more processing units, etc.; multiple elements refer to two or more elements, etc.

Parallel computing is widely used in business scenarios such as high performance computing (HPC) and artificial intelligence (AI). The message passing interface (MPI) is a set of specifications or interfaces for message passing between nodes such as CPUs and servers that participate in parallel computing. MPI supports point-to-point (P2P) communication mode and collective communication (CC) mode. Among them, point-to-point communication supports communication between a pair of processes, while group communication sets a specified process group, and all processes in the group participate in global data processing and communication operations. One process is carried on a node such as a CPU or server. MPI builds multiple different application program interfaces (APIs) based on different parallel computing demand models. These MPI interfaces complete data movement, aggregation or synchronization through different communication modes.

For example, an HPC or AI application that uses MPI for message passing, in group communication mode, usually first specifies a group of related processes to establish an MPI communication domain. The processes in the communication domain jointly implement all or part of the application functions of the HPC or AI.

Figure 1 shows a coding process for establishing an MPI communication domain. As shown in Figure 1, the coding process for establishing communication domains comm1 and comm2 is divided into the following steps: the first step is to specify communication domain members, which include all or some processes of the system specified by themselves, for example, by calling the function MPI_Group_incl() to specify the processes in group1 in the system, or calling the function MPI_Group_excl() to exclude the processes in group2 in the system; the second step is to create a communication domain, which is implemented by calling the function MPI_Comm_create(); the third step is to send/receive multicast data, which is implemented by calling an MPI interface MPI_Bcast().

Exemplary, commonly used MPI interfaces include MPI_Bcast(), MPI_Allreduce(), MPI_Scatter(), MPI_Reduce_scatter(), etc. In these typical MPI APIs, P2MP is widely used for communication.

FIG2 shows a typical MPI interface using the P2MP communication model. As shown in FIG2 , the execution system builds an MPI communication domain containing four processes according to the request of the application. The four processes run on the hardware devices GPU0-GPU3 respectively. Next, “GPU0-GPU3” is used to refer to the four corresponding members in the communication domain, or nodes, or processes for explanation.

As shown in Figure 2(a), the application execution device uses the MPI_Bcast interface and adopts the P2MP communication model of 1 sender n receivers (1SnR) in the MPI communication domain to realize data movement. As the source node of data forwarding, the other three communication members GPU1-GPU3 serve as receiving nodes. The source node GPU0 sends the received and saved data "A" to the three receiving nodes GPU0-GPU3 in P2MP mode, and the receiving nodes GPU1-GPU3 save the received data "A". At this point, the hardware devices GPU0~GPU3 can perform different processing based on the received data "A" according to the allocation instructions of the execution device to complete the calculation task of data "A" in parallel.

As shown in Figure 2(b), the execution system of the application uses the MPI_Allreduce() interface and adopts the model of multiple senders and multiple receivers (n senders n receivers, nSnR) to achieve data aggregation and movement, which can be decomposed into two steps: the first step is that a single node collects data, using the model of multiple senders and single receivers (n senders 1 receiver, nS1R); the second step is that the single node processes the collected data and sends it to multiple receiving nodes in the communication domain, using the 1SnR P2MP communication model. Specifically, the communication members GPU0/GPU1/GPU2/GPU3 first receive or establish different data "A/B/C/D" respectively. In the first step, the nS1R model is used to select the communication members GPU1/GPU2/GPU3 in the communication domain as the source nodes for data forwarding, and collect the data of multiple groups of communication members to the communication member GPU0. For example, first select the communication node GPU1 as the source node and the communication member GPU0 as the receiving node. At this time, the source node GPU1 sends the saved data "B" to GPU0 in the P2P communication mode, and the process of GPU2 and GPU3 sending data is similar. In the second step, when the communication member GPU0 completes the aggregation of the data "A", "B", "C", and "D" and performs related processing to obtain the data "A+B+C+D", it uses the P2MP method to send it. Specifically, select the communication member GPU0 as the source node, and send the data "A+B+C+D" to all other communication members GPU1-GPU3 in the P2MP communication domain in the P2MP mode. At this point, the hardware devices GPU0-GPU3 can perform different processing based on the same data "A+B+C+D" according to the system's allocation instructions to complete the calculation task of the data "A+B+C+D" in parallel.

Exemplarily, FIG3 shows a flowchart of network layer multicast forwarding and application layer multicast forwarding of P2MP data. There are multiple ways to implement the transmission of P2MP data, and a typical implementation method is to use network layer multicast or application layer multicast forwarding. As shown in FIG3(a), when using network layer multicast forwarding, the multicast distribution network will build a multicast distribution tree (MDT) for each node in the P2MP domain that serves as a multicast source. The network equipment in each MDT is responsible for forwarding P2MP data from the source node to each receiving node, that is, the sink node. As shown in FIG3(b), when using application layer multicast forwarding, P2MP data will be disassembled into multiple P2P data, and the source node 1 will send P2P data multiple times, and the P2P data will be unicast forwarded to each receiving node 2-5. When the application layer multicast forwarding is performed, some receiving nodes may participate in the forwarding. For example, after receiving the P2P data, the receiving node 2 forwards it to other receiving nodes 2-3 again.

It can be understood that in the communication domain using the P2MP communication model, regardless of the sending method (a) or (b) in Figure 3 above, for business scenarios such as HPC or AI, P2MP data needs to be reliably sent from the source node to the relevant receiving node, that is, a reliable transmission guarantee mechanism needs to be established to manage the sending/receiving of P2MP data.

Figure 4 shows the significant characteristics of P2MP data forwarding in a P2MP communication domain. As shown in Figures 4(a), (b), and (c), different nodes may be selected as the source of P2MP data transmission at different times in the life cycle of the P2MP communication domain, that is, there is frequent multicast source switching in the P2MP communication domain; in addition, the P2MP data sent in the P2MP communication domain needs to be received by all other nodes except the source node.

FIG5 shows a description of the problem of reliable forwarding of P2MP data. When the reliable transmission mechanism is established, each receiving node in the P2MP communication domain will send an ACK/NAK response message to the source node to indicate whether the receiving node has correctly received the P2MP data. Therefore, the source node of the P2MP communication domain needs to maintain the connection with each receiving node on the control plane, and also needs to manage the receiving status of each receiving node and retransmit the P2MP data to the receiving node that sent the NAK response message. When the node scale in the P2MP communication domain increases, the resource consumption of the source node is very large. As shown in FIG5(a) and (b), at time 1 and time 2 of the life cycle of the P2MP communication domain, the multicast data source in the P2MP communication domain will switch. It can be understood that any node in the P2MP communication domain may serve as a P2MP source node, and the resource consumption problem of the source node will extend to any node in the P2MP communication domain.

In the following description of this application, when each node in a distributed system performs parallel computing, the communication process between each node using remote direct memory access (RDMA) technology is taken as an example to explain the hardware configuration, software design, etc. of the distributed system for parallel computing. It is understandable that if other data access technologies (such as traditional direct connection storage, etc.) are used to exchange data between each node in the distributed system, similar hardware configuration, software design, etc. are also required in the distributed system to realize the parallel computing of business programs. In practical applications, RDMA technology directly transfers data to the storage area of the computer through the network, and quickly moves data from one system to the remote system memory. The basic communication unit of RDMA is the queue pair (QP), which can be deployed in smart network cards, channel adapters, etc. The service types based on QP include reliable transmission connection (RC), unreliable datagram connection (UD), etc.

In a possible solution, a business model is established based on business scenarios such as HPC or AI, and a P2MP communication model is used between the MPI interfaces of parallel processing nodes such as CPUs and servers. In order to ensure the reliability of P2MP data forwarding, different reliable guarantee mechanisms can be adopted according to the scale of collective communication in the business system. For example, in a small-scale collective communication scenario, the system has a P2MP data source node and each An RC connection at the transport layer is established between the receiving nodes. The RC connection is similar to a TCP connection. In this case, P2MP data uses application layer multicast technology to forward unicast data on the RC connection. At the same time, the RC connection at the transport layer implements packet loss retransmission to ensure reliable transmission of P2MP data. Alternatively, in large-scale collective communication scenarios, the system establishes a UD connection at the transport layer between the source node of the P2MP data and each receiving node to save QP resources. The UD connection is very similar to the UDP connection. In this case, P2MP uses network layer multicast or application layer multicast technology to multicast/unicast forward data packets on the UD connection. The reliable transmission of P2MP data requires the application layer of the receiving node to perform packet loss identification and retransmission requests, and the source node to perform packet loss retransmission and other operations to complete.

It can be seen that in the communication domain adopting the P2MP communication model, for the reliable guarantee mechanism of RC or UD connection, it is necessary for the control plane to establish a response mechanism, data verification mechanism and order preservation mechanism for the source node and each receiving node at the transport layer or application layer to ensure the correct forwarding of P2MP data.

Exemplarily, in order to solve the scalability problem caused by the large scale of communication nodes in the P2MP communication domain or the frequent switching of P2MP data sources, the following solutions can be adopted: establish RC connections on demand, that is, create connections between source node communication members and receiving node communication members only when there is a P2MP communication demand, thereby saving QP resources of hardware devices, but this solution will greatly affect the transmission efficiency of P2MP data because the establishment of RC connections is time-consuming; or in order to ensure low latency overhead for establishing RC connections after the source node is switched, establish RC connections in advance. Since it is impossible to determine the source node that will send P2MP data, for a P2MP communication domain containing N nodes, if a Full-Mesh connection form is adopted, N*(N-1)÷2 RC connections need to be established in advance. RC connections use limited hardware resources on hardware devices, and member processes of a communication domain may be concentrated on a certain PC host or server, which will cause RC connections to be enriched on a certain physical terminal, exhaust resources, and ultimately lead to reliability cannot be guaranteed.

In view of this, the embodiment of the present application provides a reliable transmission method for P2MP data. Based on the characteristic that each receiving node in the P2MP communication domain receives the same data, a logical interconnection network is established between multiple nodes in the P2MP communication domain, and a part of the tasks of reliable transmission processing performed by the source node is distributed and deployed to multiple nodes in the P2MP communication domain, which can effectively reduce the pressure of single-point resources and avoid single-point processing bottlenecks. In addition, by adopting a method in which multiple nodes jointly maintain reliable transmission, there is no need to establish a Full-Mesh connection between each node in the P2MP communication domain, thereby saving the cost of reliable transmission guarantee of the control plane caused by frequent switching of the source node.

When an effective response mechanism, data verification mechanism and order preservation mechanism are implemented, the reliable transmission of P2MP data within the P2MP domain can be guaranteed. In the P2MP communication domain, each receiving node receives the same data, so it is only necessary to establish an effective connection between each node at the signaling level, and design a reliable transmission scheme to implement a response mechanism (including sending a response message), a data verification mechanism and an order preservation mechanism (including packet loss request, data retransmission). The reliable transmission method provided in the embodiment of the present application decouples P2MP data forwarding and reliable transmission confirmation of P2MP data. A physical interconnection network is established between all nodes in the P2MP communication domain, that is, an interconnection network constructed by network devices, to realize the forwarding of P2MP data, and a logical interconnection network is established between all nodes in the P2MP communication domain to determine the reliable transmission confirmation relationship of P2MP data.

Exemplarily, FIG6a shows a physical internet network established between nodes in a P2MP communication domain provided by an embodiment of the present application. Specifically, the P2MP communication domain includes eight nodes, each of which has its own identification information, respectively identified as nodes 1-8. Based on the spine network architecture Spine-Leaf, a physical internet network with a secondary network topology is established between the eight nodes. P2MP data sent by any node can be forwarded in the Spine-Leaf network and reach the other seven nodes in the P2MP communication domain. FIG6b shows a forwarding path of P2MP data in a physical internet network provided by an embodiment of the present application. Based on the physical internet network shown in FIG6a, FIG6b shows a P2MP data forwarding path with node 1 as the multicast source. Node 1 can use network layer multicast (source node sends P2MP data once) to send data and forward it along the forwarding path of FIG6b.

Exemplarily, FIG7a shows a logical interconnection network established between nodes in a P2MP communication domain provided by an embodiment of the present application. Specifically, based on the P2MP communication domain and the physical interconnection network shown in FIG6a, a logical interconnection network is established between eight nodes. In the logical interconnection network, except for the first and last nodes, all nodes are directly connected to two other nodes. Two nodes with a direct connection relationship are neighbor nodes, and each node can have multiple neighbor nodes. In the P2MP communication domain, if the receiving node correctly receives the P2MP data, it needs to send an ACK response message to its neighbor node to confirm that the P2MP data has been correctly received; if the receiving node identifies that the received P2MP data is incorrect, it needs to send a NAK response message to its neighbor node to inform that the P2MP data was not received correctly, and the neighbor node retransmits the P2MP data. Figure 7b shows a reliable transmission confirmation relationship of P2MP data in a logical interconnected network provided by an embodiment of the present application. Based on the logical interconnected network shown in Figure 7a, Figure 7b shows that after node 1 sends P2MP data as a multicast source, each receiving node needs to send a response message to a neighboring node after correctly receiving the P2MP data or identifying that the P2MP data is erroneous. For example, after node 6 identifies that the P2MP data is erroneous, it needs to send a NAK response message to node 5, and node 5 completes the retransmission of the P2MP data to node 6. As a result, the task of managing the reliable transmission of P2MP data for receiving node 6, which was originally the responsibility of source node 1, is distributed to node 5, which can effectively reduce the pressure on single-point resources and avoid single-point processing bottlenecks. It can be understood that since the P2MP data received by each receiving node in the P2MP communication domain is the same, Therefore, node 5 can replace source node 1 to retransmit P2MP data to receiving node 6. In addition, in the logical interconnection network shown in 7a, it is not necessary for all receiving nodes 2-7 in the P2MP communication domain to establish a connection with source node 1, thereby saving the overhead of reliable transmission guarantee of the control plane caused by frequent switching of source nodes.

In this embodiment, the system establishes a parallel computing demand model according to the execution requirements of the application program, and allocates different CPUs or server nodes to participate in parallel computing. Based on different demand models, MPI forms a communication domain based on the P2MP communication model for all or part of the tasks. The communication domain contains multiple nodes, each of which is carried by a certain computing processing unit. These computing processing units can be located in the same or different computers, servers, clusters, storage devices, including smart network cards, channel adapters, etc. The computing processing units carried by these communication members can be physically linked through gateways, routers, etc., that is, a physical interconnection network is built between the members in the P2MP communication domain to realize the forwarding of P2MP data. On this basis, in order to realize the reliable transmission guarantee mechanism of P2MP data, it is necessary to establish a logical interconnection network between the members in the P2MP communication domain to determine the reliable transmission confirmation relationship of P2MP data.

In the P2MP communication domain, any P2MP data sent by a multicast source can be forwarded to all other receiving nodes through the physical interconnection network. The physical interconnection network can use the network layer multicast mechanism or the application layer multicast mechanism to forward P2MP data, which is selected based on the requirements of different application scenarios. The logical interconnection network establishes the adjacency relationship between the nodes in the P2MP communication domain. Based on the adjacency relationship, each node in the P2MP domain confirms the reliable transmission of the P2MP data received from the physical interconnection network to the neighboring nodes in its logical interconnection network, and the neighboring nodes are responsible for retransmitting the P2MP data.

In one example, as shown in FIG6a, there are eight nodes in the P2MP communication domain. It is understandable that the logical interconnection network between the eight nodes can have a variety of different networking modes in addition to the networking mode shown in FIG7a, so that different connections between the eight nodes can be formed. The establishment of connections between multiple nodes in the P2MP communication domain will be constrained by different strategies, for example, the number of connections between all nodes must be minimized to save resources required for establishing connections, the physical distance of the connection must be minimized so that neighboring nodes can quickly notify and respond to each other, etc. Therefore, it is necessary to build a logical interconnection network between multiple nodes in the P2MP communication domain based on different strategies and goals.

If a logical interconnection network is established based on a quick response strategy, multiple nodes can be divided into multiple node groups according to the physical distance, the physical distance between the nodes in each node group is less than or equal to the preset distance, and a connection is established between the nodes in each node group, and a node in one node group is connected to a node in another node group, thereby realizing the interconnection of multiple nodes. FIG8 shows a logical interconnection network established based on a quick response strategy provided by an embodiment of the present application, assuming that there is a certain physical neighbor relationship between certain nodes, and the physical neighbor relationship can be the relationship between nodes in a basic physical design unit (point of delivery, PoD) of a data center. PoD includes servers, access networks, converged network cabinets and their supporting facilities, which is an area of the entire network. The entire network includes multiple PoDs, for example, nodes 1-3 are deployed in the same PoD, nodes 4-6 are deployed in the second PoD, and nodes 7-8 are deployed in the third PoD. Based on the fast response strategy, first establish connections between nodes in the same PoD, and then select a node from each PoD, such as nodes 3, 6, and 7, to establish connections between the three PoDs. This maximizes the use of the physical adjacency deployment characteristics of the nodes, and reliable transmission control signaling can be processed nearby and respond quickly. However, the connections of specific nodes 3, 6, and 7 responsible for the connection between PoDs will increase. Therefore, it is necessary to avoid a node taking on more connection roles in multiple P2MP communication domains to avoid the enrichment of the number of connections on a certain node. For example, if there is a fourth PoD in the P2MP communication domain, including nodes 9-12, and its connection relationship is shown in Figure 8, when selecting node 10 to connect with nodes in other PoDs, try to connect node 10 to nodes 3 and 7 instead of node 6, so as to avoid too many connection tasks being gathered on node 6 and causing a single-point processing bottleneck.

If a logical interconnection network is established based on the control connection number strategy, each of the multiple nodes in the P2MP communication domain establishes a direct connection with a neighbor node, and no direct connection is established between the neighbor nodes of each of the multiple nodes. For example, a chain topology can be used to connect all nodes in the P2MP communication domain, so that each node needs to establish connections with at most two neighbor nodes respectively, and the total number of connections is minimized, and the total number of connections in the P2MP communication domain is evenly distributed. As shown in FIG7a, a logical interconnection network is established between eight nodes based on the control connection number strategy, and the total number of connections between the eight nodes is minimized, which will not be repeated here.

At this point, a communication connection (CC) is established between two neighboring nodes in the logical interconnection network. This communication connection is used to transmit reliable transmission response messages between neighboring nodes. The communication connection between neighboring nodes is created at the beginning of the establishment of the logical interconnection network and is used throughout the life cycle of the P2MP communication domain.

Based on the contents shown in FIGS. 6-8 , a reliable transmission method of P2MP data provided in an embodiment of the present application is introduced. It can be understood that the method can be executed by any device, equipment, platform, or device cluster with computing and processing capabilities.

For example, FIG9 shows a flow chart of a reliable transmission method for P2MP data provided by an embodiment of the present application, assuming that P2MP communication The domain includes at least a first node, a second node, and a third node. A logical interconnection network is established between the multiple nodes in the P2MP communication domain. In the logical interconnection network, the first node and the third node are neighbor nodes, and the second node and the third node are neighbor nodes. As shown in FIG9 , at a certain moment in the life cycle of the P2MP communication domain, when the second node is used as a source node for sending P2MP data, the first node and the third node are used as receiving nodes, and a confirmation process for reliable transmission of P2MP data is performed in the P2MP communication domain, including the following steps S901-S905:

Step S901: The second node sends a P2MP data packet in the P2MP communication domain. The P2MP data packet at least includes P2MP data, a sequence number of the P2MP data, and identification information of the second node (source node).

In step S902, the first node receives the P2MP data packet sent by the second node, and determines the identification information of the third node according to the forwarding table of the first node. It is assumed here that the first node determines that the received P2MP data is erroneous, and sends a response message to the third node. The response message includes at least the sequence number of the P2MP data, the status of receiving the P2MP data, the identification information of the first node, and the identification information of the second node (source node). At this time, the status flag of receiving the P2MP data is set to NAK, indicating that the received P2MP data is erroneous.

Step S903: The third node receives the P2MP data sent by the second node, and determines the identification information of the second node according to the forwarding table of the third node. It is assumed here that the third node determines that the received P2MP data is correct, and sends a response message to the second node. The response message includes at least the sequence number of the P2MP data, the status of receiving the P2MP data, the identification information of the third node, and the identification information of the second node (source node). At this time, the status flag of receiving the P2MP data is set to ACK, indicating that the P2MP data is correctly received.

It should be noted that there is no order relationship between step S902 and step S903, and the actual transmission situation depends on the physical interconnection network established by multiple nodes in the P2MP communication domain.

Step S904: The third node receives the NAK response message sent by the first node, updates the state of the first node receiving the P2MP data in the management table, and retransmits the correct P2MP data to the first node.

Step S905: The first node receives the P2MP data retransmitted by the third node, and determines the identification information of the third node according to the forwarding table of the first node. It is assumed here that the first node determines that the received P2MP data is correct, and sends a response message to the third node. The response message includes at least the sequence number of the P2MP data, the status of receiving the P2MP data, the identification information of the first node, and the identification information of the second node (source node). At this time, the status flag of receiving the P2MP data is set to ACK, indicating that the P2MP data is correctly received.

In the above steps S901-S905, each node in the P2MP communication domain has a forwarding table and a management table. The forwarding table is used to confirm a neighbor node that sends a response message after receiving a P2MP data packet sent by the source node, and the management table is used to record the state information of the managed receiving node receiving P2MP data after receiving a response message sent by the managed receiving node.

It is understandable that the P2MP data confirmation flowchart shown in FIG. 9 only lists one possible reliable transmission processing flow, and there are many other possible reliable transmission processing flows in the P2MP communication domain.

Exemplarily, Figure 10 shows a flow chart of a reliable transmission method of P2MP data provided in an embodiment of the present application, which is applied to a first node, where the first node is a node in a P2MP communication domain. A P2MP communication domain includes multiple nodes, and the first node is one of the multiple nodes. A logical interconnection network is built between the multiple nodes, and each node has its own identification information.

As shown in FIG. 10 , the reliable transmission method of P2MP data includes the following steps S1010 - S1030 .

Step S1010: receiving a P2MP data packet sent by a second node in a P2MP communication domain; the P2MP data packet at least includes P2MP data, and the second node is one of the plurality of nodes.

Step S1020, determining identification information of the third node according to the forwarding table of the first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node with a direct connection relationship in the logical interconnection network.

Step S1030: Send a response message to the third node according to the state of receiving the P2MP data and the identification information of the third node. The response message includes ACK and NAK messages. ACK indicates that the P2MP data is correctly received, and NAK indicates that the received P2MP data is erroneous.

In one example, the P2MP data packet includes at least the P2MP data and the sequence number of the P2MP data; sending a response message to the third node according to the state of receiving the P2MP data and the identification information of the third node includes: when the state of receiving the P2MP data is erroneous, sending a response message to the third node; the response message includes at least the identification information of the first node, the sequence number of the P2MP data and the state flag of receiving the P2MP data, the state flag of receiving the P2MP data is set to NAK, and NAK indicates that the received P2MP data is erroneous; the method also includes: receiving a correct P2MP data packet retransmitted by the third node.

In one example, as shown in FIG7b, the P2MP communication domain includes eight nodes, and a logical interconnection network is built between the eight nodes. Each node has its own identification information. For ease of description, node 1 is regarded as the second node, node 5 is regarded as the first node, and node 6 is regarded as the third node to illustrate the reliable transmission method of P2MP data. At a specific moment in the life cycle of the P2MP communication domain, node 1 is selected as the multicast source for sending P2MP data in the P2MP communication domain, and nodes 2-8 are selected as receiving nodes for receiving P2MP data. It can be understood that in the P2MP communication domain, any of the eight nodes may be used as the source node for sending P2MP data. In order to establish a reliable transmission mechanism, Before sending P2MP data, some additional information needs to be added to form a P2MP data packet, such as the identification information of the source node, the sequence number of the P2MP data, etc.

In the logical interconnection network shown in FIG7b, node 6 has two neighbor nodes, namely node 5 and node 7. When node 6 receives a P2MP data packet sent by node 1 in the P2MP communication domain, it determines the identification information of node 5 according to the forwarding table, and then sends a response message to node 5 according to the status of receiving the P2MP data. The response message at least includes the sequence number of the P2MP data, the status of receiving the P2MP data, the identification information of node 6, and the identification information of node 1. At this time, the status flag of receiving the P2MP data is set to ACK, indicating that the P2MP data is correctly received, or is set to NAK, indicating that the received P2MP data is incorrect. When the sent response message is NAK, it is also necessary to receive the correct P2MP data retransmitted by node 5.

In actual application, for the first node, a response message may be sent to the third node once after all P2MP data packets forwarded by the second node are received, or a response message may be sent to the third node once each time the second node receives a P2MP data packet or a fixed number of times. The selection of the above response message sending method is set in the control plane of the system. It can be understood that a fixed response message sending method can be set throughout the life cycle of the P2MP communication domain, or the response message sending method can be switched in real time according to the network congestion situation. For the convenience of description, in this application, a response message is sent once each time a P2MP data packet is received. The implementation methods of other message sending methods are similar and will not be repeated here.

FIG11 shows a logical interconnection network and a forwarding table provided in an embodiment of the present application. As shown in FIG11( a), the P2MP communication domain includes eight nodes, and a logical interconnection network is established between the nodes. A node may have multiple neighboring nodes. Therefore, after the logical interconnection network is established, it is necessary to establish a forwarding table for each node, select a neighboring node according to the forwarding table, and perform reliable transmission control signaling (ACK/NAK response message) sending and retransmission processing.

In an example, the establishment of the forwarding table mainly considers two aspects:

First, since the source node forwarding P2MP data will frequently switch during the life cycle of the P2MP communication domain, the impact of the source node switching can be considered, and the source node identification information can be used as an index item to establish a forwarding table, and the corresponding neighbor node can be selected according to different source nodes. There are two cases: for example, for node 8 in the logical interconnection network shown in Figure 11 (a), there is only one neighbor node 7, so no matter which of the nodes 1-7 is used as the source node, node 7 needs to be selected as the neighbor node for reliable transmission signaling interaction. In another case, for example, for node 6, there are two neighbor nodes 5 and 7, then the forwarding table can be established according to the identification information of different source nodes as index items.

Of course, the influence of the source node switching may not be considered, and a fixed neighbor node may be designated to interact with the reliable transmission signaling when establishing the node forwarding table. It is understandable that for all nodes in the P2MP communication domain, the influence of the source node switching may be considered when establishing the forwarding table for some of the nodes, and the identification information of the source node may be used as part of the forwarding table, while the influence of the source node switching may not be considered when establishing the forwarding table for the other nodes, and a fixed neighbor node may be designated for these nodes to interact with the reliable transmission signaling, and the influence of the source node switching may be considered or not considered for all the nodes.

Regardless of the above situation, for any node in P2MP, its forwarding table can be established using the field {key: value}, where the key in the field is the identification information or wildcard of the source node, and the value in the field is the identification information of the neighboring node for reliable transmission signaling interaction. The wildcard indicates that the impact of the source node switching is not considered, and the identification information of any source node will hit the wildcard.

Secondly, after establishing a logical interconnection network among multiple nodes in the P2MP communication domain, the establishment of a forwarding table also needs to ensure that neighboring nodes (including source nodes) that interact with each other in reliable transmission signaling can achieve effective connection on the reply message sending path.

As shown in Figure 11(b), the forwarding tables of node 2 and node 3 are given. The settings of the two forwarding tables both consider the impact of source node switching. In the forwarding table of node 2, the neighbor node identification information of node 2 is listed when the source node is node 1. Usually, in the forwarding table of a node, if the impact of source node switching is considered, it is necessary to exhaustively enumerate all nodes in the P2MP communication domain in the key item and select the corresponding neighbor node to fill in the value item. Figure 11(b) shows a schematic diagram of reliable transmission signaling interaction when node 1 is the source node and receiving nodes 2-8. It can be seen from Figure 11(b) that nodes 1-7, as neighbor nodes of other receiving nodes, have the task of processing signaling interaction, and nodes 1-7 can achieve effective connection on the path of sending the reply message based on the neighbor information configured in the forwarding table, so that nodes 1-7 can correctly receive P2MP data, thereby replacing source node 1 to achieve retransmission of P2MP data to the corresponding receiving node.

As shown in Figure 11(c), the forwarding tables of node 2 and node 3 are given. The settings of the two forwarding tables both take into account the impact of source node switching. Figure 11(c) shows a schematic diagram of reliable transmission signaling interaction when node 1 is the source node and receiving nodes 2-8. It can be seen from Figure 11(c) that nodes 1 and 3-7, as neighbor nodes of other receiving nodes, have the task of processing signaling interaction, while nodes 1 and 3-7 do not achieve effective connection on the reply message sending path under the configuration of the forwarding table, forming independent islands in two logical interconnected networks. Therefore, when nodes 3 and 4 do not correctly receive the P2MP data sent by node 1 due to network failure, they cannot obtain the correct retransmission. The P2MP packet cannot retransmit the correct P2MP data to the receiving node it manages, so it cannot guarantee that each receiving node in the P2MP communication domain can correctly receive the P2MP data.

As shown in Figure 11(d), the forwarding tables of nodes 2 and 3 are given. The setting of the forwarding table of node 3 takes into account the influence of the source node switching, while the setting of the forwarding table of node 2 does not take into account the influence of the source node switching. The neighboring nodes to which node 2 sends the reply message are wildcarded. Figure 11(d) shows a schematic diagram of reliable transmission signaling interaction when node 1 is used as the source node and receiving nodes 2-8. As can be seen from Figure 11(c), nodes 2-7, as neighboring nodes of other receiving nodes, have the task of processing signaling interaction, while nodes 2-7 are not connected to node 1 on the reply message path under the configuration of the forwarding table, forming independent islands in two logical interconnected networks. Therefore, when nodes 2 and 3 do not correctly receive the P2MP data sent by node 1 due to network failure, they cannot obtain the retransmitted correct P2MP packets, nor can they retransmit the correct P2MP data to the receiving nodes they manage, thereby failing to ensure that each receiving node in the P2MP communication domain can correctly receive the P2MP data.

Therefore, the forwarding tables of node 2 and node 3 shown in Figure 11(c) and Figure 11(d) cannot meet the requirements of the forwarding table settings in this application and need to be modified accordingly.

Therefore, by establishing a logical interconnection network for multiple nodes in the P2MP communication domain, the task of reliable transmission and processing of P2MP data performed only by the source node is distributed and deployed to multiple nodes in the P2MP communication domain, which can effectively reduce the pressure of single-point processing.

Exemplarily, Figure 12 shows a flow chart of a reliable transmission method of P2MP data provided in an embodiment of the present application, which is applied to a third node, where the third node is a node in a P2MP communication domain. A P2MP communication domain includes multiple nodes, and the third node is one of the multiple nodes. A logical interconnection network is built between the multiple nodes, and each node has its own identification information.

As shown in FIG. 12 , the reliable transmission method of P2MP data includes the following steps S1210 - S1220 .

Step S1210, receiving a response message from the first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node with a direct connection relationship in the logical interconnection network; the response message includes at least identification information of the first node and a status flag of the first node receiving P2MP data; the P2MP data is included in a P2MP data packet, the P2MP data packet is sent by the second node in a P2MP communication domain, and the second node is one of the multiple nodes.

Step S1220: updating the management table of the third node according to the response message of the first node; the management table at least includes the state of the first node receiving the P2MP data.

In an example, the third node updates the status of the first node receiving the P2MP data in the management table according to the response message of the first node.

In one example, after receiving a response message from the first node, the reliable transmission method of P2MP data further includes: determining whether the status flag of the first node receiving P2MP data in the response message of the first node is NAK, and if so, retransmitting the correct P2MP data packet to the first node and starting a timer; wherein NAK indicates that the P2MP data received by the first node is incorrect; receiving the response message from the first node again, and stopping the timer. Similarly, when the third node receives the response message from the first node again, it is also necessary to update the status of the first node receiving P2MP data in the management table.

In one example, in a P2MP communication domain, the third node (a receiving node) and the second node (source node) as neighbor nodes of other receiving nodes have the task of processing signaling interaction. For the second node, after sending the P2MP data, it is necessary to start a timer for all receiving nodes that need to send a reply message to the second node. After receiving the reply message from each receiving node corresponding to the timer, the timer started for this receiving node is closed. If the timer times out, the correct P2MP data is retransmitted to this receiving node. For the third node, it is necessary to perform similar operations for all receiving nodes that need to send a reply message to the third node after receiving the P2MP data, which will not be repeated here.

After multiple nodes in the P2MP communication domain establish a logical interconnection network, it is necessary to establish a management table for each node. Usually, it is necessary to consider the impact of the source node switching, and use the identification information of the source node as an index item to establish the management table. Of course, the impact of the source node switching may also be ignored. It can be understood that the establishment principle here is similar to the description of the "first aspect" in the process of establishing the forwarding table in step S1030, and will not be repeated here.

For any node in P2MP, such as the third node, its management table can be established using the field {key: value: state}, where the key in the field is the identification information or wildcard of the source node sending the P2MP data, the value in the field is the identification information of the corresponding receiving node, and the state in the field is the state flag of the corresponding receiving node receiving the P2MP data.

For any node in P2MP, it is necessary to maintain a forwarding table and a management table. The two tables can be two independent tables or combined into one table. It is only necessary to ensure that when looking up the table, the identification information of the neighbor node that sends the response message and the receiving node that receives the response message can be obtained respectively.

Therefore, by establishing a logical interconnection network between communication nodes in the P2MP communication domain, the mechanism for implementing reliable transmission of multicast data is distributed and deployed to multiple communication nodes in the P2MP communication domain, which can effectively alleviate the problem of reliable transmission from only the source node to multiple receiving nodes. Processing pressure and avoid single-point processing bottlenecks.

Exemplarily, FIG. 13a shows a network layer multicast resource configuration diagram based on RC connection in a P2MP communication domain provided by an embodiment of the present application. A P2MP communication domain includes seven communication nodes, each of which can be carried in the same or different computers, servers, clusters, storage devices, including computing processing units such as smart network cards and channel adapters, which can realize physical layer links through gateways, routers, etc. At a certain moment in the life cycle of the P2MP communication domain, node 1 is selected as the source node for sending P2MP data. Based on the logical interconnected network established by each node in the P2MP communication domain in Figure 13a, the network layer multicast resource configuration based on the RC connection is set between the source node 1 and each receiving node 2-7 through the control plane. It can be seen from Figure 13a that the number of RC connections that need to be established is 6, among which node 1 establishes 1 RC connection with receiving nodes 2, 3, and 4 respectively, and nodes 2, 4, and 5 establish 1 RC connection with receiving nodes 5, 7, and 6 respectively. Analogously to the establishment of a Full-Mesh connection among seven communication nodes, the number of RC connections that need to be established is 7*(7-1)÷2=21, thereby saving the overhead of reliable transmission guarantee of the control plane caused by frequent switching of the source node.

Exemplarily, Figure 13b shows a logical connection diagram based on the data plane within a P2MP communication domain provided by an embodiment of the present application. As shown in Figure 13b, the source node 1 makes multiple copies of the P2MP data in the multicast network and then sends them to the receiving nodes 2-7 at one time.

Exemplarily, FIG. 13c shows a schematic diagram of a logical connection based on a control plane in a P2MP communication domain provided by an embodiment of the present application. As shown in FIG13c, there are the following types of nodes in the P2MP communication domain, for example, node 1 (multicast source) → node 2 (multicast sink 1, a receiving node managed by the multicast source, a neighbor node of multicast sink 2) → node 5 (multicast sink 2, a receiving node managed by multicast sink 1, a neighbor node of multicast sink 3) → node 6 (multicast sink 3, a receiving node managed by multicast sink 2). In their respective reliable transmission confirmation domains, node 2 (multicast sink 1): first, according to the forwarding table, it is determined that the neighbor node in the reliable transmission control domain is node 1. Secondly, it is necessary to send a message to the multicast source whether the node can reliably receive P2MP data, and it is not necessary to inform the multicast source whether the node can reliably receive P2MP data. Whether the receiving node groupcast sink 2 managed by the multicast source itself can reliably receive P2MP data, node 5 (multicast sink 2): first determine the neighbor node in the reliable transmission control domain as node 2 according to the forwarding table, and then need to notify its neighbor node groupcast sink 1 whether it can reliably receive P2MP data, and do not need to send whether this node can reliably receive P2MP data to the multicast source, and do not need to inform the receiving node groupcast sink 3 managed by the multicast source whether it can reliably receive P2MP data, node 6 (multicast sink 3): first determine the neighbor node in the reliable transmission control domain as node 5 according to the forwarding table, and then need to notify its neighbor node groupcast sink 2 whether it can reliably receive P2MP data, and do not need to send whether this node can reliably receive P2MP data to the multicast source. The working principle of the reliable transmission guarantee mechanism of other nodes 3, 4, and 7 is similar and will not be repeated. In the logical interconnection network, multicast sink 1 can correctly receive the P2MP data sent by the multicast source, and can also retransmit the correct P2MP data when multicast sink 2 sends a NAK response message. In addition, any node only needs to feedback information to its neighboring node, which may be a multicast source or a multicast sink. Therefore, through the distributed reliable transmission control domain division, the pressure of a single point can be effectively reduced and the bottleneck of single point processing can be avoided.

Based on the method in the above embodiment, exemplarily, Figure 14 shows a schematic diagram of a reliable transmission device for P2MP data provided in an embodiment of the present application. The device can be deployed on a first node. The first node is a node in a P2MP communication domain. A P2MP communication domain includes multiple nodes. The first node is one of the multiple nodes. A logical interconnection network is built between the multiple nodes. Each node has its own identification information. As shown in Figure 14, the confirmation device 1400 includes: a communication module 1410 and a processing module 1420.

The communication module 1410 may receive a P2MP data packet sent by a second node in a P2MP communication domain; the P2MP data packet includes at least P2MP data, and the second node is one of the plurality of nodes.

The processing module 1420 can determine the identification information of the third node according to the forwarding table of the first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node with a direct connection relationship in the logical interconnection network; the processing module 1420 can also send a response message to the third node according to the status of receiving P2MP data and the identification information of the third node.

In some embodiments, the P2MP data packet includes at least the P2MP data and the sequence number of the P2MP data; the processing module 1420 may send a response message to the third node according to the state of receiving the P2MP data and the identification information of the third node, and specifically, when the state of receiving the P2MP data is erroneous, send a response message to the third node; the response message includes at least the identification information of the first node, the sequence number of the P2MP data and the state flag of receiving the P2MP data, and the state flag of receiving the P2MP data is set to NAK, which indicates that the received P2MP data is erroneous; the communication module 1410 may also receive a correct P2MP data packet retransmitted by the third node.

In some embodiments, the P2MP data packet includes at least P2MP data and identification information of the second node; the processing module 1420 can determine the identification information of the third node based on the forwarding table of the first node, specifically, determine the identification information of the third node based on the forwarding table of the first node and the identification information of the second node; wherein the forwarding table of the first node includes at least one field {key: value}, the key in the field is the identification information of the second node or a wildcard, and the value in the field is the identification information of the third node.

In some embodiments, a logical interconnection network is established by connecting multiple nodes, and the connections between the multiple nodes include: each of the multiple nodes has at least one neighbor node, each of the multiple nodes establishes a direct connection with the neighbor node, and no direct connection is established between the neighbor nodes of each of the multiple nodes.

In some embodiments, a logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including: multiple nodes are divided into multiple node groups, each node group includes a first node, the physical distance between the nodes in the node group is less than or equal to a preset distance, connections are established between the nodes in the node group, and connections are established with the first nodes in the multiple node groups.

Exemplarily, Figure 15 shows a schematic diagram of a reliable transmission device for P2MP data provided in an embodiment of the present application. The device can be deployed at a third node. The third node is a node in a P2MP communication domain. A P2MP communication domain includes multiple nodes. The third node is one of the multiple nodes. A logical interconnection network is built between the multiple nodes. Each node has its own identification information. As shown in Figure 15, the confirmation device 1500 includes: a communication module 1510 and a processing module 1520.

The communication module 1510 can receive a response message from the first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node with a direct connection relationship in the logical interconnection network; the response message includes at least identification information of the first node and a status flag of the first node receiving P2MP data; the P2MP data is included in a P2MP data packet, the P2MP data packet is sent by the second node in a P2MP communication domain, and the second node is one of multiple nodes.

The processing module 1520 may update the management table of the third node according to the response message of the first node; the management table at least includes the state of the first node receiving the P2MP data.

In some embodiments, after receiving the response message from the first node, the processing module 1520 can determine whether the status flag of the first node receiving the P2MP data in the response message of the first node is NAK. If so, retransmit the correct P2MP data packet to the first node and start the timer; wherein NAK indicates that the P2MP data received by the first node is incorrect; the communication module 1510 is also used to: receive the response message from the first node and turn off the timer.

In some embodiments, the P2MP data packet includes at least P2MP data and identification information of the second node; the processing module 1520 can update the management table of the second node according to the response message of the first node, and specifically, update the state flag of the first node receiving P2MP data in the management table of the second node according to the identification information of the first node and the state flag of the first node receiving P2MP data; wherein the management table of the second node includes at least one field {key: value: state}, the key in the field is the identification information or a wildcard of the second node, the value in the field is the identification information of the first node, and the state in the field is the state flag of the first node receiving P2MP data.

In some embodiments, a logical interconnection network is established by connecting multiple nodes, and the connections between the multiple nodes include: each of the multiple nodes has at least one neighbor node, each of the multiple nodes establishes a direct connection with the neighbor node, and no direct connection is established between the neighbor nodes of each of the multiple nodes.

In some embodiments, a logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including: multiple nodes are divided into multiple node groups, each node group includes a third node, the physical distance between the nodes in the node group is less than or equal to a preset distance, the nodes in the node group are connected, and the third nodes in multiple node groups are connected.

Based on the method in the above embodiment, an embodiment of the present application provides an electronic device. The electronic device may include: a display screen; at least one memory for storing programs; at least one processor for executing the programs stored in the memory. Wherein, when the program stored in the memory is executed, the processor is used to execute the method described in the above embodiment. Exemplarily, the electronic device may be a mobile phone, a tablet computer, a desktop computer, a laptop computer, a handheld computer, a notebook computer, a server, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook, and a cellular phone, a personal digital assistant (personal digital assistant, PDA), an augmented reality (augmented reality, AR) device, a virtual reality (virtual reality, VR) device, an artificial intelligence (artificial intelligence, AI) device, a wearable device, a vehicle-mounted device, a smart home device and/or a smart city device. The embodiment of the present application does not impose any special restrictions on the specific type of the electronic device.

Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the method in the above embodiment.

Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the method in the above embodiment.

It is understood that the processor in the embodiments of the present application may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.

In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions.

Claims (23)

1. A reliable transmission method for P2MP data, characterized in that it is applied to a first node, the first node is a node in a P2MP communication domain, the P2MP communication domain includes multiple nodes, the first node is one of the multiple nodes, a logical interconnection network is established between the multiple nodes, and each node has its own identification information, the method comprising:

   receiving a P2MP data packet sent by a second node in the one P2MP communication domain; the P2MP data packet at least includes P2MP data, and the second node is one of the plurality of nodes;

   Determine identification information of a third node according to a forwarding table of the first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node having a direct connection relationship in the logical interconnection network;

   A response message is sent to the third node according to the state of receiving the P2MP data and the identification information of the third node.
2. The method according to claim 1, wherein the P2MP data packet at least includes P2MP data and a sequence number of the P2MP data;

   The sending a response message to the third node according to the state of receiving the P2MP data and the identification information of the third node includes:

   When the state of receiving the P2MP data is erroneous, sending a response message to the third node; the response message includes at least the identification information of the first node, the sequence number of the P2MP data, and a state flag of receiving the P2MP data, the state flag of receiving the P2MP data is set to NAK, and the NAK indicates that the received P2MP data is erroneous;

   The method further includes: receiving a correct P2MP data packet retransmitted by the third node.
3. The method according to claim 1, wherein the P2MP data packet at least includes P2MP data and identification information of the second node;

   The determining the identification information of the third node according to the forwarding table of the first node includes:

   Determine the identification information of the third node based on the forwarding table of the first node and the identification information of the second node; wherein the forwarding table of the first node contains at least one field {key: value}, the key in the field is the identification information or a wildcard of the second node, and the value in the field is the identification information of the third node.
4. The method according to claim 1, wherein the logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including:

   Each of the multiple nodes has at least one neighbor node, each of the multiple nodes establishes a direct connection with the neighbor node, and the neighbor nodes of each of the multiple nodes do not establish a direct connection with each other.
5. The method according to claim 1, wherein the logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including:

   The multiple nodes are divided into multiple node groups, each node group includes a first node, the physical distance between the nodes in the node group is less than or equal to a preset distance, connections are established between the nodes in the node group, and the first node in the multiple node groups is connected.
6. A reliable transmission method for P2MP data, characterized in that it is applied to a third node, the third node is a node in a P2MP communication domain, the P2MP communication domain includes multiple nodes, the third node is one of the multiple nodes, a logical interconnection network is established between the multiple nodes, and each node has its own identification information, the method comprising:

   receiving a response message from a first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node having a direct connection relationship in the logical interconnection network; the response message includes at least identification information of the first node and a state flag of the first node receiving P2MP data; the P2MP data is included in a P2MP data packet, and the P2MP data packet is sent by a second node in the one P2MP communication domain, and the second node is one of the multiple nodes;

   According to the response message of the first node, a management table of the third node is updated; the management table at least includes a state in which the first node receives the P2MP data.
7. The method according to claim 6, characterized in that after receiving the response message from the first node, the method further comprises:

   Determine whether a status flag indicating that the first node receives the P2MP data in a response message of the first node is NAK, and if so, retransmit a correct P2MP data packet to the first node and start a timer; wherein the NAK indicates that the P2MP data received by the first node is incorrect;

   A response message is received from the first node, and the timer is turned off.
8. The method according to claim 6, wherein the P2MP data packet at least includes P2MP data and identification information of the second node;

   The updating the management table of the third node according to the response message of the first node includes:

   According to the identification information of the first node and the state flag of the first node receiving the P2MP data, the state flag of the first node receiving the P2MP data in the management table of the third node is updated; wherein the management table of the second node includes at least one field {key: value: state}, the key in the field is the identification information or a wildcard of the second node, the value in the field is the identification information of the first node, and the state in the field is the state flag of the first node receiving the P2MP data.
9. The method according to claim 6, wherein the logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including:

   Each of the multiple nodes has at least one neighbor node, each of the multiple nodes establishes a direct connection with the neighbor node, and the neighbor nodes of each of the multiple nodes do not establish a direct connection with each other.
10. The method according to claim 6, wherein the logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected, including:

    The multiple nodes are divided into multiple node groups, each node group includes a third node, the physical distance between the nodes in the node group is less than or equal to a preset distance, the nodes in the node group are connected, and the third nodes in the multiple node groups are connected.
11. A reliable transmission device for P2MP data, characterized in that it is deployed on a first node, the first node is a node in a P2MP communication domain, the P2MP communication domain includes multiple nodes, the first node is one of the multiple nodes, a logical interconnection network is established between the multiple nodes, each node has its own identification information, and the device includes:

    a communication module, configured to receive a P2MP data packet sent by a second node in the one P2MP communication domain; the P2MP data packet at least includes P2MP data, and the second node is one of the plurality of nodes;

    A processing module, configured to determine identification information of a third node according to a forwarding table of the first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node having a direct connection relationship in the logical interconnection network;

    The processing module is further configured to send a response message to the third node according to the state of receiving the P2MP data and the identification information of the third node.
12. The device according to claim 11, wherein the P2MP data packet at least includes P2MP data and a sequence number of the P2MP data;

    When the processing module sends a response message to the third node according to the state of receiving the P2MP data and the identification information of the third node, it is used to:

    When the state of receiving the P2MP data is erroneous, sending a response message to the third node; the response message includes at least the identification information of the first node, the sequence number of the P2MP data, and a state flag of receiving the P2MP data, the state flag of receiving the P2MP data is set to NAK, and the NAK indicates that the received P2MP data is erroneous;

    The communication module is further used for receiving the correct P2MP data packet retransmitted by the third node.
13. The device according to claim 11, wherein the P2MP data packet at least includes P2MP data and identification information of the second node;

    When the processing module determines the identification information of the third node according to the forwarding table of the first node, it is used to:

    Determine the identification information of the third node based on the forwarding table of the first node and the identification information of the second node; wherein the forwarding table of the first node contains at least one field {key: value}, the key in the field is the identification information or a wildcard of the second node, and the value in the field is the identification information of the third node.
14. The device according to claim 11, wherein the logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected including:

    Each of the multiple nodes has at least one neighbor node, each of the multiple nodes establishes a direct connection with the neighbor node, and the neighbor nodes of each of the multiple nodes do not establish a direct connection with each other.
15. The device according to claim 11, wherein the logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected including:

    The multiple nodes are divided into multiple node groups, each node group includes a first node, the physical distance between the nodes in the node group is less than or equal to a preset distance, connections are established between the nodes in the node group, and the first node in the multiple node groups is connected.
16. A reliable transmission device for P2MP data, characterized in that it is deployed on a third node, the third node is a node in a P2MP communication domain, the P2MP communication domain includes multiple nodes, the third node is one of the multiple nodes, a logical interconnection network is established between the multiple nodes, each node has its own identification information, and the device includes:

    a communication module, configured to receive a response message from a first node; in the logical interconnection network, the first node has at least one neighbor node, and the third node is one of the neighbor nodes of the first node; the neighbor node is a node having a direct connection relationship in the logical interconnection network; the response message includes at least identification information of the first node and a state flag of the first node receiving P2MP data; the P2MP data is included in a P2MP data packet, and the P2MP data packet is sent by a second node in the one P2MP communication domain, and the second node is one of the multiple nodes;

    The processing module is used to update the management table of the third node according to the response message of the first node; the management table at least includes the state of the first node receiving the P2MP data.
17. The device according to claim 16, characterized in that after the processing module receives the response message from the first node, it is further used to:

    Determine whether a status flag indicating that the first node receives the P2MP data in a response message of the first node is NAK, and if so, retransmit a correct P2MP data packet to the first node and start a timer; wherein the NAK indicates that the P2MP data received by the first node is incorrect;

    The communication module is further used for: receiving a response message from the first node and shutting down the timer.
18. The apparatus according to claim 16, wherein the P2MP data packet comprises at least P2MP data and identification information of the second node;

    When the processing module updates the management table of the second node according to the response message of the first node, it is used to:

    According to the identification information of the first node and the state flag of the first node receiving the P2MP data, the state flag of the first node receiving the P2MP data in the management table of the second node is updated; wherein the management table of the second node includes at least one field {key: value: state}, the key in the field is the identification information or a wildcard of the second node, the value in the field is the identification information of the first node, and the state in the field is the state flag of the first node receiving the P2MP data.
19. The device according to claim 16, wherein the logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected including:

    Each of the multiple nodes has at least one neighbor node, each of the multiple nodes establishes a direct connection with the neighbor node, and the neighbor nodes of each of the multiple nodes do not establish a direct connection with each other.
20. The device according to claim 16, wherein the logical interconnection network is established by connecting multiple nodes, and the multiple nodes are connected including:

    The plurality of nodes are divided into a plurality of node groups, each node group includes a third node, and the physical distance between the nodes in the node group is less than the equal distance between the nodes. At a preset distance, the nodes in the node group are connected, and the third nodes in the plurality of node groups are connected.
21. An electronic device, comprising:

    at least one memory for storing a program;

    at least one processor, configured to execute the program stored in the memory;

    Wherein, when the program stored in the memory is executed, the processor is used to execute the method according to any one of claims 1-10.
22. A computer-readable storage medium stores a computer program, and when the computer program runs on a processor, the processor executes the method according to any one of claims 1 to 10.
23. A computer program product, characterized in that when the computer program product runs on a processor, the processor is caused to execute the method according to any one of claims 1 to 10.