Network flow path distribution method and device, electronic equipment and storage medium

By using feature extraction, message delivery and path allocation ratio mapping in computing power network, the path allocation of network flow is optimized, the network congestion problem caused by ignoring topological structure and traffic relationship in the prior art is solved, and network performance is improved and reliability.

CN119052197AActive Publication Date: 2024-11-29PENG CHENG LAB
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411003043.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-29
Estimated Expiration
2044-07-24

AI Technical Summary

Technical Problem

Under the computing power network architecture, existing machine learning methods often ignore the relationship between topology and traffic when solving multipath transmission problems, resulting in network congestion and performance degradation.

Method used

A method of path allocation for network flow is proposed. Through the path allocation model of feature extraction layer, message delivery aggregation layer and readout layer, the traffic demand and path capacity matrix are obtained, and feature extraction, message delivery update and path allocation ratio mapping are carried out. Consider the relationship between network flows and other network flows and optimize path allocation.

Benefits of technology

The network is adaptively optimized traffic allocation is realized, network performance is improved, and the risk of network congestion is reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119052197A_ABST
    Figure CN119052197A_ABST
Patent Text Reader

Abstract

The invention discloses a network flow path allocation method and device, electronic equipment and a storage medium, and the method comprises the steps: inputting a flow demand quantity and a path capacity matrix into a path allocation model, enabling the path allocation model to fully pay attention to a pre-configured transmission path, namely a topological structure, during the processing, obtaining a predicted path allocation ratio through the path allocation model, and obtaining a path allocation result. And when it is detected that the predicted path distribution ratio distribution meets a preset constraint condition, taking the predicted path distribution ratio as target path distribution ratio distribution of the network flow. According to the method and the device, the topological structure of network flow transmission and the relationship with other network flows are combined, and the obtained target path distribution ratio distribution can realize adaptive optimization flow distribution of the network, so that the overall performance of the network is improved, and the risk of network congestion is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of network multi-path transmission, and in particular to a method, device, electronic device and storage medium for allocating paths of network flows. Background Art

[0002] With the emergence of computing network architecture, centralized machine learning has become a powerful tool to solve the problem of multi-path transmission. The core idea of ​​SDN is to decouple the control plane and data plane of the network and introduce a centralized controller to effectively manage the entire network. This architecture enhances the flexibility and programmability of the network. The centralized control architecture of the computing network provides a unified interface for centralized machine learning, making it easier to integrate into network management. This centralized management mode eliminates the bottleneck of decentralized control in traditional networks and creates favorable conditions for the introduction of intelligent solutions. With the widespread adoption of SDN, network administrators can intuitively monitor and control the network and obtain real-time traffic and device status information. This provides a richer and more real-time data source for machine learning algorithms, enabling them to more accurately model network status, predict traffic changes, and adjust network configuration when necessary.

[0003] However, in the current machine learning methods for solving multi-path transmission problems under the computing power network architecture, the topology structure, that is, the transmission path of the traffic, is often ignored, and the relationship between the traffic and other traffic is not taken into account, which may cause network congestion and performance degradation. Summary of the invention

[0004] The present application aims to solve at least one of the technical problems existing in the prior art. To this end, the present application proposes a network flow path allocation method, device, electronic device and storage medium, which can improve network performance and reduce the risk of network congestion.

[0005] To achieve the above-mentioned purpose, the first aspect of the present application provides a method for allocating paths of network flows, characterized in that a multi-path scheduling controller is applied to a computing power network system for multi-path transmission, wherein the computing power network system for multi-path transmission further comprises a plurality of network nodes, each of the network nodes is connected to each other, and the multi-path controller is connected to each of the network nodes respectively;

[0006] The method comprises:

[0007] Get the traffic demand of network flow;

[0008] Determine a preset number of preconfigured transmission paths based on a source node and a destination node of the network flow; wherein the source node and the destination node are each one of the plurality of network nodes;

[0009] constructing a path capacity matrix of the preconfigured transmission path;

[0010] Inputting the traffic demand and the path capacity matrix into a path allocation model; the path allocation model is provided with a feature extraction layer, a message transmission aggregation layer and a readout layer;

[0011] Performing feature extraction processing on the flow demand and the path capacity matrix through the feature extraction layer to obtain a first network flow feature;

[0012] Performing message passing update processing on the first network flow feature through the message passing layer to obtain a second network flow feature;

[0013] Mapping the second network flow feature through the readout layer to obtain a predicted path allocation ratio distribution of the network flow on each of the preconfigured transmission paths;

[0014] It is detected whether the predicted path allocation ratio meets a preset constraint condition, and when it is detected that the predicted path allocation ratio distribution meets the preset constraint condition, the predicted path allocation ratio is used as the target path allocation ratio distribution of the network flow.

[0015] According to some embodiments of the present application, constructing a path capacity matrix of the preconfigured transmission path includes:

[0016] Determining the link through which each of the preconfigured transmission paths passes;

[0017] Obtain the link capacity of the passed link;

[0018] The path capacity matrix is ​​constructed based on the link capacity.

[0019] According to some embodiments of the present application, performing message passing update processing on the first network flow feature through the message passing layer to obtain the second network flow feature includes:

[0020] Based on the route links of the preconfigured transmission path of the network flow, determine an adjacent network flow adjacent to the network flow; wherein the adjacent network flow has at least one route link that is the same as the network flow;

[0021] Performing message passing processing on the first network flow feature and each first adjacent network flow feature in sequence to obtain a plurality of message passing features; wherein the first adjacent network flow feature is obtained by the feature extraction layer performing feature extraction processing on the adjacent network flow;

[0022] Aggregate the plurality of message transmission features to obtain an aggregate feature;

[0023] Updating the aggregated features based on a gated recurrent unit to obtain updated features;

[0024] Repeat the above steps until the number of iterations reaches a preset iteration threshold, and use the updated feature finally obtained as the second network flow feature.

[0025] According to some embodiments of the present application, mapping the second network flow feature through the readout layer to obtain the predicted path allocation ratio distribution of the network flow on each of the preconfigured transmission paths includes:

[0026] By mapping and associating the second network flow feature with a preset number of the preconfigured transmission paths respectively, a preset number of associated features are obtained;

[0027] The associated features are input into a reading function to obtain the predicted path allocation ratio distribution of the network flow on each of the preconfigured transmission paths.

[0028] According to some embodiments of the present application, the preset constraint condition includes a first constraint condition, and the first constraint condition is:

[0029]

[0030] Among them, σ s,d (p) represents the allocation ratio of the preconfigured transmission path P; s represents the source node of the network flow; d represents the destination node of the network flow; p represents the preconfigured transmission path; P s,d represents the set of preconfigured transmission paths from the source node s to the destination node d; V represents the set of network nodes.

[0031] According to some embodiments of the present application, the preset constraint condition includes a second constraint condition, and the second constraint condition is:

[0032]

[0033] Among them, σ s,d (p) represents the allocation ratio of the preconfigured transmission path P; s represents the source node of the network flow; d represents the destination node of the network flow; p represents the preconfigured transmission path; P s,d represents the set of preconfigured transmission paths from the source node s to the destination node d; V represents the set of network nodes.

[0034] According to some embodiments of the present application, the preset constraint condition includes a third constraint condition, and the third constraint condition is:

[0035]

[0036] Among them, σ s,d(p) represents the allocation ratio of the preconfigured transmission path P; s represents the source node of the network flow; d represents the destination node of the network flow; p represents the preconfigured transmission path; P s,d represents the set of preconfigured transmission paths from the source node s to the destination node d; f represents the flow demand of the network flow; F s,d represents the set of traffic demand of the network flow from the source node to the destination node d; c(e) represents the link capacity of the passing links; E represents the set of the passing links.

[0037] To achieve the above-mentioned purpose, the second aspect of the present application provides a network flow path allocation device, which is applied to a multi-path scheduling controller in a computing power network system with multi-path transmission, wherein the computing power network system with multi-path transmission also includes a plurality of network nodes, each of which is connected to each other, and the multi-path controller is connected to each of the network nodes respectively;

[0038] The device comprises:

[0039] An acquisition module is used to obtain the traffic demand of the network flow;

[0040] A determination module, configured to determine a preset number of preconfigured transmission paths based on a source node and a destination node of the network flow; wherein the source node and the destination node are each one of the plurality of network nodes;

[0041] A construction module, used to construct a path capacity matrix of the preconfigured transmission path;

[0042] An input module, used for inputting the traffic demand and the path capacity matrix into a path allocation model; the path allocation model is provided with a feature extraction layer, a message transmission aggregation layer and a readout layer;

[0043] A feature extraction module, used for performing feature extraction processing on the flow demand and the path capacity matrix through the feature extraction layer to obtain a first network flow feature;

[0044] A message passing update module, used for performing message passing update processing on the first network flow feature through the message passing layer to obtain a second network flow feature;

[0045] A mapping module, configured to perform mapping processing on the second network flow feature through the readout layer to obtain a predicted path allocation ratio distribution of the network flow on each of the preconfigured transmission paths;

[0046] The detection module is used to detect whether the predicted path allocation ratio meets the preset constraint condition. When it is detected that the predicted path allocation ratio distribution meets the preset constraint condition, the predicted path allocation ratio is used as the target path allocation ratio distribution of the network flow.

[0047] To achieve the above-mentioned purpose, an embodiment of the third aspect of the present application provides an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the path allocation method for the network flow described in any one of the embodiments of the first aspect.

[0048] To achieve the above-mentioned purpose, the fourth aspect of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the path allocation method for network flow described in any one of the first aspect embodiments.

[0049] The path allocation method, device, electronic device and storage medium of the network flow of the present application input the traffic demand and path capacity matrix into the path allocation model, so that the path allocation model can fully pay attention to the pre-configured transmission path, that is, the topological structure, when processing, and then in the path allocation model, the traffic demand and path capacity matrix are subjected to feature extraction processing by the feature extraction layer to obtain the first network flow feature, and the first network flow feature is subjected to message transmission update processing by the message transmission layer to obtain the second network flow feature. In the process of message transmission update processing, the information exchange between the network flow and other network flows is realized, so that when the predicted path allocation ratio distribution is obtained by the readout layer, the relationship between the network flow and other network flows is fully considered, and then when it is detected that the predicted path allocation ratio distribution meets the preset constraint conditions, the predicted path allocation ratio is used as the target path allocation ratio distribution of the network flow. In this way, the path allocation method of the network flow of the present application combines the topological structure of the network flow transmission and the relationship with other network flows, and the obtained target path allocation ratio distribution can realize the adaptive optimization of network traffic allocation, thereby improving the overall performance of the network and reducing the risk of network congestion.

[0050] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The present application is further described below with reference to the accompanying drawings and embodiments, wherein:

[0052] Figure 1 A schematic diagram of the structure of a computing power network system for multi-path transmission according to an embodiment of the present application;

[0053] Figure 2A schematic diagram of the steps of the method for allocating paths for network flows according to an embodiment of the present application;

[0054] Figure 3 A schematic diagram of the processing process of the path allocation model of an embodiment of the present application;

[0055] Figure 4 for Figure 2 A specific flow chart of step S230;

[0056] Figure 5 for Figure 2 A specific flow chart of step S260;

[0057] Figure 6 for Figure 2 A specific flow chart of step S270;

[0058] Figure 7 A schematic diagram of the structure of a path allocation device for a network flow according to an embodiment of the present application;

[0059] Figure 8 A schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0060] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as limiting the present application.

[0061] In the description of the present application, it should be understood that descriptions involving orientation, such as up, down, front, back, left, right, etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0062] In the description of this application, "several" means more than one, "more" means more than two, "greater than", "less than", "exceed", etc. are understood to exclude the number itself, and "above", "below", "within", etc. are understood to include the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0063] In the description of this application, unless otherwise clearly defined, terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in this application based on the specific content of the technical solution.

[0064] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0065] First, some nouns involved in this application are analyzed:

[0066] Inband Network Telemetry (INT) is a hybrid measurement technology that uses data plane service traffic to collect, carry, organize, and report network conditions, rather than using a separate control plane management traffic to collect the above information. INT technology usually includes three main functional nodes: INT Source (source node), INTSink (aggregation node), and INT Transit Hop (transmission node). INT Source: Responsible for indicating the traffic and type of information to be collected. When the data packet reaches the INT Source, an INT header is added with an instruction set (INT Instruction) indicating the type of information to be collected. INT Transit Hop: All devices on the line that support INT telemetry. When the message is forwarded to these nodes, the device collects the corresponding information (INTMetadata) according to the instruction set in the INT header and inserts it into the INT message. INT Sink: Responsible for organizing and reporting the received information to the monitoring device. When the message reaches the INT Sink, all INT information will pop up and be forwarded to the telemetry server through an appropriate method (such as gRPC).

[0067] Message Passing Neural Network (MPNN): A graph neural network model based on the message passing mechanism. Its core idea is to update the feature representation of nodes by passing messages between nodes in the graph. This message passing process is essentially a local-to-global aggregation process, in which the local information of each node is modeled through a neural network and then integrated with the global information. The MPNN framework consists of two main steps: message passing and node updating. Message passing: In this step, each node receives information from its neighbor nodes and constructs a message. This message can be calculated based on the features of the node and its neighbor nodes. All messages passed from neighbor nodes are aggregated to obtain the aggregate message of the current node. The aggregation method can be summation, averaging or other forms of weighted aggregation. Node updating: During the node updating process, MPNN updates the representation of the node based on the node's own features and the aggregated message. Usually, the node's feature representation and the aggregated message are concatenated or added to obtain a new feature representation. Then, this new feature representation is processed through a neural network module (such as a fully connected layer) to obtain the updated feature representation of the node.

[0068] The Gate Recurrent Unit (GRU) is a variant of the Recurrent Neural Network (RNN) used to process sequence data. Compared with traditional RNN, GRU performs better in capturing long-term dependencies in sequence data, and has higher efficiency and faster training speed. GRU solves the gradient vanishing and gradient exploding problems that traditional RNNs are prone to when processing long sequences by introducing two key mechanisms: Update Gate and Reset Gate. These two gating mechanisms allow the GRU unit to selectively retain and update information at each time step of the sequence, thereby effectively capturing long-term dependencies.

[0069] The centralized control architecture of the computing network provides a unified interface for centralized machine learning, making it easier to integrate into network management. This centralized management model eliminates the bottleneck of decentralized control in traditional networks and creates favorable conditions for the introduction of intelligent solutions. With the widespread adoption of SDN, network administrators can intuitively monitor and control the network and obtain real-time traffic and device status information. This provides a richer and more real-time data source for machine learning algorithms, enabling them to more accurately model network status, predict traffic changes, and adjust network configurations when necessary.

[0070] However, current machine learning methods for solving multipath transmission problems in computing network architectures have an important oversight: the topology is often ignored, and other attributes and topologies of the traffic (connectivity information between nodes), such as the capacity of the links that the traffic needs to pass through, the number of end-to-end hops, etc., are not considered, which makes these methods unable to effectively deal with the challenges brought by link failures. This can lead to network congestion, performance degradation, and failure to meet user expectations for changing computing business needs. For example, when a specific computing demand traffic suddenly increases, the demand for other traffic that shares a common link with the computing demand traffic will be affected accordingly. This can lead to poor performance when traffic fluctuates significantly.

[0071] Based on this, the present application proposes a network flow path allocation method, device, electronic device and storage medium, which can improve network performance and reduce the risk of network congestion.

[0072] The path allocation method for network flow in the embodiments of the present application can be applied to a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a router, a programmable switch, a network card, etc.; the server can be configured as an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the path allocation method for network flow, etc., but is not limited to the above forms.

[0073] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific services or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which services are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0074] Reference Figure 1 , Figure 1This is a schematic diagram of the structure of the computing power network system with multi-path transmission of the embodiment of the present application. The computing power network system with multi-path transmission includes a multi-way scheduling controller, a network information collector, multiple computing power clusters and multiple switches. Multiple computing power clusters are respectively connected to a switch, and each switch is connected to multiple other switches to form a computing power network. Each switch serves as a network node. Usually, the source node and the destination node are switches directly connected to the computing power cluster. The network information collector is connected to each switch respectively, the multi-way scheduling controller is connected to each switch respectively, and the multi-way scheduling controller is connected to the network information collector. The network information collector is used to collect data information in the switch. The multi-way scheduling controller is used to execute the path allocation method of the network flow of the embodiment of the present application.

[0075] Reference Figure 2 ,based on Figure 1 In a computing network system for multi-path transmission, the first embodiment of the present application proposes a method for allocating paths for network flows. The method for allocating paths for network flows is applied to Figure 1 A multi-path scheduling controller in a computing network system with multi-path transmission. Figure 2 The illustrated method flow includes but is not limited to steps S210 to S280.

[0076] Step S210, obtaining the traffic demand of the network flow;

[0077] It is worth noting that in the network, a network flow usually refers to a series of data packets transmitted from a source IP address and port combination to a destination IP address and port combination within a period of time. These data packets have the same five-tuple information, namely the source IP address, destination IP address, source port number, destination port number and transport layer protocol (such as TCP or UDP). Flow is an abstract description of this orderly, continuous and directional data transmission process.

[0078] It is worth noting that the Figure 1 The network information collector in the embodiment determines the traffic demand of the network flow based on the in-band network telemetry method. The traffic demand refers to the process in which the network flow is transmitted in the link. The link needs to have a certain capacity to transmit the network flow. The capacity requirement of the link is the traffic demand of the network flow. In one embodiment, the traffic demand of the network flow is measured in bits per second (bps) or the number of packets transmitted per second.

[0079] Step S220, determining a preset number of preconfigured transmission paths based on a source node and a destination node of the network flow; wherein the source node and the destination node are each one of the plurality of network nodes;

[0080] It is worth noting that for network flows, there are multiple transmission paths from the source node to the destination node. The present application determines a preset number of transmission paths from multiple transmission paths as pre-configured transmission paths, which can reduce the complexity of the subsequently constructed traffic demand matrix and improve processing efficiency.

[0081] In one embodiment, the preconfigured transmission path is determined by the following steps:

[0082] Determine all transmission paths from the source node to the destination node;

[0083] Determine the number of links passed by each transmission path; wherein two adjacent network nodes constitute a link, a passing link refers to a link passed by the transmission path, and each transmission path includes at least one passing link.

[0084] The transmission paths are sorted in ascending order of the number of links, and the transmission paths with a preset number of links ranked first are obtained as pre-configured paths.

[0085] This embodiment obtains a preconfigured transmission path through the above steps, and the preconfigured transmission path has fewer links. Therefore, when transmitting through the preconfigured transmission path, the transmission delay will be shorter, thereby improving transmission efficiency and improving network performance.

[0086] Step S230, constructing a path capacity matrix of the preconfigured transmission path;

[0087] It is worth noting that the path capacity matrix of the transmission path includes the information of the preconfigured transmission path, so that when the subsequent path allocation model makes predictions, it can simultaneously focus on the traffic demand of the network flow and the preconfigured transmission path, and can optimize the prediction results of the path allocation model.

[0088] Step S240, inputting the traffic demand and the path capacity matrix into a path allocation model; the path allocation model is provided with a feature extraction layer, a message transmission aggregation layer and a readout layer;

[0089] Step S250, performing feature extraction processing on the traffic demand and the path capacity matrix through a feature extraction layer to obtain a first network flow feature;

[0090] Step S260, performing message transmission update processing on the first network flow feature through the message transmission layer to obtain a second network flow feature;

[0091] Step S270, mapping the second network flow feature through the readout layer to obtain the predicted path allocation ratio distribution of the network flow on each preconfigured transmission path;

[0092] It is worth noting that referring to Figure 3 , Figure 3The figure is a schematic diagram of the processing process of the path allocation model. The path allocation model is based on a message passing neural network. The input layer of the message passing neural network is used as the feature extraction layer of the path allocation model, and the traffic demand and path traffic matrix are converted into processable neural network inputs. The feature extraction layer performs feature extraction processing on the traffic demand and path capacity matrix to obtain the first network flow feature. The message passing layer of the message passing network is used as the message passing aggregation layer of the path allocation model. The message passing layer performs message passing update processing on the first network flow feature to achieve comprehensive perception of the dynamic changes of the overall network topology and traffic distribution, so that the second network flow feature contains richer information and provides a more accurate reference for subsequent predictions. Then the output layer of the message passing neural network is used as the reading layer of the path allocation model. The second network flow feature is mapped in the reading layer to obtain the predicted path allocation ratio distribution of the network flow on each preconfigured transmission path.

[0093] Step S280, detecting whether the predicted path allocation ratio meets the preset constraint condition, and when it is detected that the predicted path allocation ratio distribution meets the preset constraint condition, using the predicted path allocation ratio as the target path allocation ratio distribution of the network flow.

[0094] It is worth noting that the target path allocation ratio distribution includes the allocation ratio of each preconfigured transmission path. The allocation ratio refers to the traffic proportion of the network flow. For example, if the allocation ratio of a preconfigured transmission path is 30%, then during transmission, the preconfigured transmission path is used to transmit 30% of the traffic in the network flow.

[0095] It is worth noting that if it is detected that the predicted path allocation ratio distribution does not satisfy the preset constraint condition, step S210 to step S270 are re-executed until the predicted path allocation ratio distribution satisfies the preset constraint condition.

[0096] The embodiment of the present application inputs the traffic demand and the path capacity matrix into the path allocation model through steps S210 to S280, so that the path allocation model can fully pay attention to the pre-configured transmission path, that is, the topological structure, during processing. Then, in the path allocation model, the traffic demand and the path capacity matrix are subjected to feature extraction processing by the feature extraction layer to obtain the first network flow feature, and the first network flow feature is subjected to message transmission update processing by the message transmission layer to obtain the second network flow feature. In the process of message transmission update processing, the information exchange between the network flow and other network flows is realized, so that when the predicted path allocation ratio distribution is obtained through the readout layer, the relationship between the network flow and other network flows is fully considered, and then when it is detected that the predicted path allocation ratio distribution meets the preset constraint conditions, the predicted path allocation ratio is used as the target path allocation ratio distribution of the network flow. In this way, the path allocation method of the network flow of the present application combines the topological structure of the network flow transmission and the relationship with other network flows, and the obtained target path allocation ratio distribution can realize the adaptive optimization of the network flow allocation, thereby improving the overall performance and reducing the risk of network congestion.

[0097] It should be noted that steps S210 to S280 are executed periodically. For example, steps S210 to S280 can be executed every 5 minutes. In this way, real-time monitoring of the network status and real-time path allocation of network flows are achieved, and the effect of dynamic allocation can be achieved, which can effectively improve the utilization rate of network resources.

[0098] In one embodiment, referring to Figure 4 , Figure 4 for Figure 2 A specific flow chart of step S230 in FIG. Figure 4 The illustrated steps include but are not limited to steps S410 to S430.

[0099] Step S410, determining the link through which each preconfigured transmission path passes;

[0100] It is noteworthy that the network nodes through which each preconfigured transmission path passes are determined, wherein adjacent network nodes form a passing link, thereby being able to determine the passing link through which each of the preconfigured transmission paths passes.

[0101] Step S420, obtaining the link capacity of the passing link;

[0102] It is worth noting that after determining the link, Figure 1 The network information collector in determines the link capacity of the traversing links based on the in-band network telemetry method.

[0103] Step S430: construct a path capacity matrix based on the link capacity.

[0104] It is worth noting that the embodiment of the present application also constructs a flow demand matrix based on the flow demand. Figure 3 , Figure 3 The structure of the traffic demand matrix and the structure of the path capacity matrix are illustrated. The traffic demand matrix has only one column, and each element is a traffic demand. Each preconfigured transmission path corresponds to a path capacity matrix, and a traffic demand corresponds to a preset number of path capacity matrices. There is only one row in the path capacity matrix. In the path capacity matrix, each element represents a link. When the value of the element is not 0, it means that the link is a passing link, and the value of the element is the link capacity of the passing link; when the value of the element is 0, it means that the link is not a passing link of the preconfigured transmission path.

[0105] The embodiment of the present application constructs a path capacity matrix through steps S410 to S430. In step S250, the feature extraction layer splices and combines a flow demand with its corresponding preconfigured transmission path to obtain a first network flow feature corresponding to the flow demand. The first network flow feature includes the flow demand of the network flow and the relevant information of a preset number of preconfigured transmission paths, so that when the subsequent path allocation model makes a prediction, it can pay attention to the flow demand of the network flow and the preconfigured transmission path at the same time, and can optimize the prediction result of the path allocation model.

[0106] In one embodiment, referring to Figure 5 , Figure 5 for Figure 2 A specific flow chart of step S260 in FIG. Figure 5 The illustrated step flow includes but is not limited to step S510 to step S550.

[0107] Step S510, determining an adjacent network flow adjacent to the network flow based on the path links of the preconfigured transmission path of the network flow; wherein the adjacent network flow has at least one path link that is the same as the network flow;

[0108] It is worth noting that there are multiple network flows in a computing network system with multi-path transmission. Determine the adjacent network flows adjacent to the current network flow from the multiple network flows. Adjacent network flows refer to other network flows that have the same path link as the current network flow. When performing message passing, the rule of message passing is that message passing occurs only when two network flows share the same link, which allows the current network flow to exchange information with flows that interact with other network flows.

[0109] Step S520, performing message passing processing on the first network flow feature and each first adjacent network flow feature in turn to obtain a plurality of message passing features; wherein the first adjacent network flow feature is obtained by performing feature extraction processing on the adjacent network flow by the feature extraction layer;

[0110] Specifically, the message passing process can be expressed as:

[0111]

[0112] Among them, Passing represents the message passing function, Characterize the jth message transmission characteristics, A first network flow feature characterizing a current network flow; Characterizes the first adjacent network flow feature of the jth adjacent network flow; t represents the time step t. Message passing processing can capture the dynamic relationship between network flows, especially the mutual influence between network flows, enhance the path allocation model's understanding of the network state, and improve the prediction accuracy of the path allocation model.

[0113] Step S530, aggregating multiple message transmission features to obtain aggregate features;

[0114] Step S540, updating the aggregated features based on the gated recurrent unit to obtain updated features;

[0115] Specifically, the update process can be expressed as:

[0116]

[0117] Among them, at this time represents the update feature, t represents the time step t, and Aggregation represents the aggregation processing function; Characterize the jth message passing feature; Update characterizes the update function. The embodiment of the present application achieves comprehensive perception of the dynamic changes of the overall network topology and traffic distribution by aggregating the message passing features, providing a more accurate reference for subsequent decision-making. The aggregation process integrates the information of each network flow, paying special attention to the dynamic relationship between the capacity of the link passed and the traffic demand. The detailed capture of this relationship helps to better understand how traffic propagates in the network. The use of gated recurrent units (GRUs) to update the aggregated features can better capture time series information, enable the path allocation model to adapt to the evolution of the network state, make full use of previous information and current message passing results, accurately capture the dynamic evolution of traffic demand in the network, and improve the ability to model time dynamics. A more comprehensive understanding of the network state can be achieved, thereby achieving more intelligent decision adjustments to meet real-time network needs.

[0118] Step S550, repeat the above steps until the number of iterations reaches a preset iteration threshold, and use the updated feature finally obtained as the second network flow feature.

[0119] It is worth noting that step S510 to step S540 are repeatedly executed until the number of iterations reaches the preset iteration threshold, and the updated feature finally obtained is used as the second network flow feature; the preset iteration threshold is T, that is, iterates T times, and As the second network flow feature, t is 1, ..., 2, ... T. It is conducive to the gradual transmission and aggregation of information and captures the dynamic relationship between flows in the network.

[0120] The embodiment of the present application realizes the message transmission between the current network flow and other network flows through steps S510 to S550, obtains the message transmission characteristics, aggregates the message transmission characteristics to obtain the aggregated characteristics, and then updates the aggregated characteristics through the gated recurrent unit (GRU) to obtain the second network flow characteristics. In this way, the path allocation model can adapt to the evolution of the network state, make full use of previous information and current message transmission results, accurately capture the dynamic evolution of traffic demand in the network, and improve the modeling ability of time dynamics. It can have a more comprehensive understanding of the network state and improve the prediction effect of the path allocation model.

[0121] In one embodiment, referring to Figure 6 , Figure 6 for Figure 2 A specific flow chart of step S270 in FIG. Figure 6 The illustrated step flow includes but is not limited to step S610 to step S620.

[0122] Step S610, mapping and associating the second network flow feature with a preset number of preconfigured transmission paths respectively, to obtain a preset number of associated features;

[0123] It is worth noting that referring to Figure 3 , a fully connected layer is provided in the reading layer, in which the second network flow feature is respectively mapped and associated with a preset number of preconfigured transmission paths to obtain a preset number of associated features. In the fully connected layer, the process of establishing the mapping association is expressed as:

[0124]

[0125] in, Characterize the second network flow characteristics, The associated feature characterizes the association between the second network flow feature and the i-th preconfigured transmission path, and fully_connected characterizes the function of the fully connected mapping, which is used to map and associate the second network flow feature with the preconfigured transmission path.

[0126] Step S620: input the associated features into the reading function to obtain the predicted path allocation ratio distribution of the network flow on each preconfigured transmission path.

[0127] It is worth noting that referring to Figure 3 In the fully connected layer, the read function uses the softmax mapping method to ensure The values ​​in are converted into probability distribution, which indicates the proportion of network flow f to preconfigured path i. The process of step S620 is expressed as:

[0128]

[0129] Among them, σ s,d (p i ) represents the allocation ratio of the preconfigured transmission path Pi, read represents the read function, A correlation feature characterizing the association between the second network flow feature and the i-th preconfigured transmission path.

[0130] In one embodiment, the preset constraint condition includes a first constraint condition, and the first constraint condition is:

[0131]

[0132] Among them, σ s,d (p) represents the allocation ratio of the preconfigured transmission path P; s represents the source node of the network flow; d represents the destination node of the network flow; p represents the preconfigured transmission path; P s,d represents the set of preconfigured transmission paths from source node s to destination node d; V represents the set of network nodes. In the first constraint, the allocation ratio of each preconfigured transmission path is restricted to be non-negative to simulate the actual network environment.

[0133] The preset constraint condition includes a second constraint condition, which is:

[0134]

[0135] Among them, σ s,d (p) represents the allocation ratio of the preconfigured transmission path P; s represents the source node of the network flow; d represents the destination node of the network flow; p represents the preconfigured transmission path; P s,d represents a set of preconfigured transmission paths from source node s to destination node d; V represents a set of network nodes. In the second constraint, the sum of the allocation ratios on the predefined paths must be equal to 1.

[0136] The preset constraint condition includes the third constraint condition, which is:

[0137]

[0138] Among them, σ s,d (p) represents the allocation ratio of the preconfigured transmission path P; s represents the source node of the network flow; d represents the destination node of the network flow; p represents the preconfigured transmission path; Ps,d represents the set of preconfigured transmission paths from source node s to destination node d; f represents the network flow; F s,d represents the set of network flows from source node d to destination node d; c(e) represents the link capacity of the passed links; E represents the set of passed links. The third constraint is used to ensure that the flow distribution is within the capacity range of the passed links.

[0139] In one embodiment, based on Figure 1 The computing power network system with multi-path transmission is trained by the following steps:

[0140] Get the training traffic demand of the training network flow; the training traffic demand is obtained in the same way as the traffic demand.

[0141] A preset number of preconfigured transmission training paths are determined based on a source node and a destination node of a training network flow; wherein the source node and the destination node are each one of a plurality of network nodes; and the method for determining the preconfigured transmission training path is the same as the method for determining the preconfigured transmission path.

[0142] A path capacity training matrix of the preconfigured transmission training path is constructed; specifically, the path capacity training matrix is ​​constructed in the same manner as the path capacity matrix.

[0143] Input the training traffic demand and path capacity training matrix into the initial path allocation model;

[0144] The feature extraction layer performs feature extraction processing on the training traffic demand and path capacity training matrix to obtain the first network flow training feature; specifically, the feature extraction processing process for the training traffic demand and path capacity training matrix is ​​the same as the feature extraction processing process for the traffic demand and path capacity matrix.

[0145] The first network flow training feature is updated through the message passing layer to obtain the second network flow training feature; specifically, the process of performing message passing update processing on the first network flow training feature is the same as the process of performing message transmission update processing on the first network flow feature.

[0146] The second network flow training features are mapped through the readout layer to obtain the training path allocation ratio distribution of the network flow on each preconfigured transmission training path; specifically, the process of mapping the second network flow training features is the same as the process of mapping the second network flow features.

[0147] Detect whether the training path allocation ratio meets the preset constraint conditions. When it is detected that the training path allocation ratio distribution meets the preset constraint conditions, calculate the loss value based on the training path allocation ratio and the loss function.

[0148] The loss function is:

[0149]

[0150] Among them, σ s,d (p) represents the allocation ratio of the preconfigured transmission training path P; s represents the source node of the network flow; d represents the destination node of the network flow; p represents the preconfigured transmission training path; P s,d represents the set of preconfigured transmission training paths from source node s to destination node d; f represents the training network flow; F s,d Characterize the set of training network flows from source node d to destination node d.

[0151] After the loss value is calculated, the model parameters of the path allocation model are optimized based on the loss value, and the above steps are repeated and iterated until the loss value is greater than the preset threshold or the loss value converges, thereby obtaining a trained path allocation model.

[0152] It is worth noting that in the path allocation model of the present application, the loss function aims to maximize network traffic. Therefore, the predicted path allocation ratio distribution obtained by the path allocation model can realize adaptive optimization of network traffic allocation, thereby improving overall performance and reducing the risk of network congestion.

[0153] Reference Figure 7 The second aspect of the present application provides a network flow path allocation device, which is applied to Figure 1 A multi-path scheduling controller in a computing network system with multi-path transmission;

[0154] The device includes:

[0155] An acquisition module 710 is used to acquire the flow demand of the network flow;

[0156] A determination module 720, configured to determine a preset number of preconfigured transmission paths based on a source node and a destination node of a network flow; wherein the source node and the destination node are each one of a plurality of network nodes;

[0157] A construction module 730, configured to construct a path capacity matrix of a preconfigured transmission path;

[0158] An input module 740 is used to input the traffic demand and path capacity matrix into the path allocation model; the path allocation model is provided with a feature extraction layer, a message transmission aggregation layer and a readout layer;

[0159] A feature extraction module 750 is used to perform feature extraction processing on the traffic demand and path capacity matrix through a feature extraction layer to obtain a first network flow feature;

[0160] A message passing update module 760, configured to perform message passing update processing on the first network flow feature through a message passing layer to obtain a second network flow feature;

[0161] A mapping module 770 is used to perform mapping processing on the second network flow feature through the readout layer to obtain a predicted path allocation ratio distribution of the network flow on each preconfigured transmission path;

[0162] The detection module 780 is used to detect whether the predicted path allocation ratio meets the preset constraint condition. When it is detected that the predicted path allocation ratio distribution meets the preset constraint condition, the predicted path allocation ratio is used as the target path allocation ratio distribution of the network flow.

[0163] The path allocation device of the network flow is used to execute the path allocation method of the network flow of the first embodiment of the present application. When executing the method, the flow demand and the path capacity matrix are input into the path allocation model, so that the path allocation model can fully pay attention to the pre-configured transmission path, that is, the topological structure, when processing, and then in the path allocation model, the flow demand and the path capacity matrix are subjected to feature extraction processing by the feature extraction layer to obtain the first network flow feature, and the first network flow feature is subjected to message transmission update processing by the message transmission layer to obtain the second network flow feature. In the process of message transmission update processing, the information exchange between the network flow and other network flows is realized, so that when the predicted path allocation ratio distribution is obtained by the readout layer, the relationship between the network flow and the other network flows is fully considered, and then when the predicted path allocation ratio distribution is detected to meet the preset constraint condition, the predicted path allocation ratio is used as the target path allocation ratio distribution of the network flow. In this way, when the path allocation device of the network flow of the present application is used, it combines the topological structure of the network flow transmission and the relationship with other network flows, and the obtained target path allocation ratio distribution can realize the adaptive optimization of network flow allocation, thereby improving the overall performance and reducing the risk of network congestion.

[0164] In one embodiment, the construction module 730 includes a first determination submodule, an acquisition submodule and a construction submodule.

[0165] The first determination submodule is used to determine the link through which each preconfigured transmission path passes;

[0166] The acquisition submodule is used to obtain the link capacity of the passing link;

[0167] The construction submodule is used to construct a path capacity matrix based on link capacity.

[0168] In one embodiment, the message passing update module 760 includes a second determination submodule, a message passing processing submodule, an aggregation submodule, an update submodule, and an iteration submodule.

[0169] The second determination submodule is used to determine an adjacent network flow adjacent to the network flow based on the path link of the preconfigured transmission path of the network flow; wherein the adjacent network flow has at least one path link that is the same as the network flow;

[0170] The message passing processing submodule is used to perform message passing processing on the first network flow feature and each first adjacent network flow feature in turn to obtain multiple message passing features; wherein the first adjacent network flow feature is obtained by the feature extraction layer performing feature extraction processing on the adjacent network flow;

[0171] The aggregation submodule is used to aggregate multiple message transmission features to obtain aggregate features;

[0172] The updating submodule is used to update the aggregated features based on the gated recurrent unit to obtain updated features;

[0173] The iterative submodule is used to repeatedly iterate the above steps until the number of iterations reaches a preset iteration threshold, and the finally obtained updated feature is used as the second network flow feature.

[0174] In one embodiment, the mapping module 770 includes an association submodule and a reading submodule.

[0175] The association submodule is used to establish mapping associations with a preset number of preconfigured transmission paths through the second network flow feature, respectively, to obtain a preset number of association features;

[0176] The reading submodule is used to input the associated features into the reading function to obtain the predicted path allocation ratio distribution of the network flow on each preconfigured transmission path.

[0177] Reference Figure 8 , Figure 8 This is a schematic diagram of the structure of an electronic device according to an embodiment of the third aspect of the present application. The electronic device includes:

[0178] The processor 801 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0179] The memory 802 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 802 can store an operating system and other applications. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 802, and the processor 801 calls and executes the path allocation method of the network flow in the embodiment of the present application;

[0180] Input / output interface 803, used to implement information input and output;

[0181] The communication interface 804 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.);

[0182] A bus 805 that transmits information between the various components of the device (e.g., the processor 801, the memory 802, the input / output interface 803, and the communication interface 804);

[0183] The processor 801 , the memory 802 , the input / output interface 803 and the communication interface 804 are connected to each other in communication within the device via a bus 805 .

[0184] The fourth aspect of the present application is a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the path allocation method for network flow of any one of the first aspect embodiments.

[0185] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0186] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0187] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0188] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0189] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0190] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0191] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the mapping relationship of the mapping objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next mapping objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0192] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0193] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0194] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0195] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0196] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A method for allocating paths of network flows, characterized in that: A multi-path scheduling controller applied to a computing power network system for multi-path transmission, wherein the computing power network system for multi-path transmission further comprises a plurality of network nodes, each of the network nodes is interconnected, and the multi-path controller is respectively connected to each of the network nodes; The method comprises: Get the traffic demand of network flow; Determine a preset number of preconfigured transmission paths based on a source node and a destination node of the network flow; wherein the source node and the destination node are each one of the plurality of network nodes; constructing a path capacity matrix of the preconfigured transmission path; Inputting the traffic demand and the path capacity matrix into a path allocation model; the path allocation model is provided with a feature extraction layer, a message transmission aggregation layer and a readout layer; Performing feature extraction processing on the flow demand and the path capacity matrix through the feature extraction layer to obtain a first network flow feature; Performing message passing update processing on the first network flow feature through the message passing layer to obtain a second network flow feature; Mapping the second network flow feature through the readout layer to obtain a predicted path allocation ratio distribution of the network flow on each of the preconfigured transmission paths; It is detected whether the predicted path allocation ratio meets a preset constraint condition, and when it is detected that the predicted path allocation ratio distribution meets the preset constraint condition, the predicted path allocation ratio is used as the target path allocation ratio distribution of the network flow.

2. The method for allocating paths of network flows according to claim 1, characterized in that: The constructing the path capacity matrix of the preconfigured transmission path includes: Determining the link through which each of the preconfigured transmission paths passes; Obtain the link capacity of the passed link; The path capacity matrix is ​​constructed based on the link capacity.

3. The method for network flow path allocation according to claim 2, characterized in that: The performing message passing update processing on the first network flow feature through the message passing layer to obtain the second network flow feature includes: Based on the route links of the preconfigured transmission path of the network flow, determine an adjacent network flow adjacent to the network flow; wherein the adjacent network flow has at least one route link that is the same as the network flow; Performing message passing processing on the first network flow feature and each first adjacent network flow feature in sequence to obtain a plurality of message passing features; wherein the first adjacent network flow feature is obtained by the feature extraction layer performing feature extraction processing on the adjacent network flow; Aggregate the plurality of message transmission features to obtain an aggregate feature; Updating the aggregated features based on a gated recurrent unit to obtain updated features; Repeat the above steps until the number of iterations reaches a preset iteration threshold, and use the updated feature finally obtained as the second network flow feature.

4. The method for allocating paths of network flows according to claim 1, characterized in that: The mapping process of the second network flow feature through the readout layer to obtain the predicted path allocation ratio distribution of the network flow on each of the preconfigured transmission paths includes: Establishing mapping associations between the second network flow features and a preset number of the preconfigured transmission paths respectively, to obtain a preset number of association features; The associated features are input into a reading function to obtain the predicted path allocation ratio distribution of the network flow on each of the preconfigured transmission paths.

5. The method for network flow path allocation according to claim 1, characterized in that: The preset constraint condition includes a first constraint condition, and the first constraint condition is: Among them, σ s,d (p) represents the allocation ratio of the preconfigured transmission path P; s represents the source node of the network flow; d represents the destination node of the network flow; p represents the preconfigured transmission path; P s,d represents the set of preconfigured transmission paths from the source node s to the destination node d; V represents the set of network nodes.

6. The network flow path allocation method according to claim 1, characterized in that: The preset constraint condition includes a second constraint condition, and the second constraint condition is: Among them, σ s,d (p) represents the allocation ratio of the preconfigured transmission path P; s represents the source node of the network flow; d represents the destination node of the network flow; p represents the preconfigured transmission path; P s,d represents the set of preconfigured transmission paths from the source node s to the destination node d; V represents the set of network nodes.

7. The network flow path allocation method according to claim 2, characterized in that: The preset constraint condition includes a third constraint condition, and the third constraint condition is: Among them, σ s,d (p) represents the allocation ratio of the preconfigured transmission path P; s represents the source node of the network flow; d represents the destination node of the network flow; p represents the preconfigured transmission path; P s,d represents the set of preconfigured transmission paths from source node s to destination node d; f represents the network flow; F s,d represents the set of network flows from source node d to destination node d; c(e) represents the link capacity of the passing links; and E represents the set of passing links.

8. A network flow path allocation device, characterized in that: A multi-path scheduling controller applied to a computing power network system for multi-path transmission, wherein the computing power network system for multi-path transmission further comprises a plurality of network nodes, each of the network nodes is interconnected, and the multi-path controller is respectively connected to each of the network nodes; The device comprises: An acquisition module is used to obtain the traffic demand of the network flow; A determination module, configured to determine a preset number of preconfigured transmission paths based on a source node and a destination node of the network flow; wherein the source node and the destination node are each one of the plurality of network nodes; A construction module, used to construct a path capacity matrix of the preconfigured transmission path; An input module, used for inputting the traffic demand and the path capacity matrix into a path allocation model; the path allocation model is provided with a feature extraction layer, a message transmission aggregation layer and a readout layer; A feature extraction module, used for performing feature extraction processing on the flow demand and the path capacity matrix through the feature extraction layer to obtain a first network flow feature; A message passing update module, used for performing message passing update processing on the first network flow feature through the message passing layer to obtain a second network flow feature; A mapping module, configured to perform mapping processing on the second network flow feature through the readout layer to obtain a predicted path allocation ratio distribution of the network flow on each of the preconfigured transmission paths; The detection module is used to detect whether the predicted path allocation ratio meets the preset constraint condition. When it is detected that the predicted path allocation ratio distribution meets the preset constraint condition, the predicted path allocation ratio is used as the target path allocation ratio distribution of the network flow.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the path allocation method for network flow according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the network flow path allocation method according to any one of claims 1 to 7 is implemented.

Citation Information

Patent Citations

  • Flow control method and device for hybrid SDN network

    CN111147387A

  • TSN network joint route selection and flow distribution method and related equipment

    CN114286413A

  • Network traffic load balancing method based on reinforcement learning

    CN116455824A

  • Methods, computing devices, and storage media for adjusting network traffic

    US20210168065A1