CN114900441A

CN114900441A - Network performance prediction method, performance prediction model training method and related device

Info

Publication number: CN114900441A
Application number: CN202210467299.6A
Authority: CN
Inventors: 崔勇; 王莫为; 惠林博; 梁铷; 刘振华
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2022-04-29
Filing date: 2022-04-29
Publication date: 2022-08-12
Anticipated expiration: 2042-04-29
Also published as: CN114900441B

Abstract

The application relates to the field of artificial intelligence, in particular to a network performance prediction method, a network performance prediction model training method and a related device. Wherein predicting comprises: acquiring a heterogeneous relationship diagram at the current moment and first state information of the heterogeneous relationship diagram at the current moment aiming at a target network; processing the heterogeneous relationship diagram at the current moment and the first state information of the heterogeneous relationship diagram at the current moment through a network performance prediction model of the target network to obtain second state information of the heterogeneous relationship diagram at the current moment, wherein the second state information comprises a performance index of the target network at the current moment; and determining the first state information of the heterogeneous relationship diagram at the current moment at the next moment based on the second state information and the first state information of the heterogeneous relationship diagram at the current moment. By adopting the embodiment of the application, not only can the performance index of the network be accurately predicted, but also the performance index changing along with time can be obtained.

Description

Network performance prediction method, performance prediction model training method and related device

Technical Field

The present application relates to the field of artificial intelligence, and in particular, to a network performance prediction method, a performance prediction model training method, and a related apparatus.

Background

Network researchers and operators both strive to improve network quality of service to meet various network service demands, often relying on network performance models to achieve efficient network design, operation, and optimization. In particular, the network performance model provides a new means for performance monitoring in online scenarios and performance optimization in offline scenarios. An operator can acquire real network performance in real time through the performance model and monitor possible abnormity; a designer can evaluate whether the performance under given topology, configuration and flow meets the design requirement through a performance model; researchers can guide the design and evaluation of optimization schemes such as traffic scheduling and routing strategies through the performance model.

With the rapid development of machine learning technology, deep learning methods based on deep neural networks are applied to the fields of computer vision, machine translation and the like and have great success. Researchers have begun focusing on data-driven approach-based network performance modeling and have made some progress, with representative work being: 1. a network performance prediction model is built based on the graph neural network to determine the relationship between the links and the paths, and further the steady-state time delay of the path levels under different topologies, routing schemes, link bandwidths and flows can be evaluated. 2. A network performance prediction model is constructed based on a variational auto-encoder (VAE) and long short-term memory (LSTM) to determine the relationship between the path delay and the flow, and further the path delay and the packet loss distribution can be predicted through a flow matrix of a time sequence under a given topology.

The two network performance prediction models lack systematic expression capability and time sequence state modeling capability. There are a variety of network entities in a real network, and there are a variety of local or global-scale configurations on these entities. For example, the cache size of a switch only affects the length of the queue locally, while routing policies affect the traffic distribution throughout the network. Network performance indicators are also shown on different network entities, such as wan scenarios focusing on path-level delay and packet loss, and data center scenarios focusing on flow-level indicators such as Flow Completion Time (FCT). Because the existing network performance prediction model lacks systematic expression capability, the performance index of the network can be predicted only based on the relationship between the link and the path, and the performance index determined in the way is inaccurate. In addition, in the process of constantly changing network performance, the state of the network at the next moment depends on the state at the current moment due to the cache of the network device and the data packet being sent, and the change process of the network performance can be accurately reflected only by continuous performance evaluation, so that some network performance abnormalities (such as time delay increase and throughput decrease) can be identified, and therefore, a model needs to output a performance index changing along with time. Because the existing network performance prediction model does not have the time sequence state modeling capability, the existing network performance prediction model cannot output the performance index changing along with time.

Disclosure of Invention

The embodiment of the application provides a network performance prediction method, a network performance prediction model training method and a related device.

In a first aspect, an embodiment of the present application provides a network performance prediction method, including:

acquiring a current heterogeneous relationship graph and first state information of the current heterogeneous relationship graph aiming at a target network, wherein the current heterogeneous relationship graph is used for representing a plurality of nodes participating in network operation in the current target network and relations among the nodes; processing the heterogeneous relationship diagram at the current moment and the first state information of the heterogeneous relationship diagram at the current moment through a network performance prediction model of the target network to obtain second state information of the heterogeneous relationship diagram at the current moment, wherein the second state information comprises a performance index of the target network at the current moment; and determining the first state information of the heterogeneous relationship diagram at the current moment at the next moment based on the second state information and the first state information of the heterogeneous relationship diagram at the current moment.

In one possible embodiment, the plurality of nodes of the target network comprises: network devices, links, queues, and paths. In some scenarios, the plurality of nodes of the target network also include a flow. The network device may include a switch and a server.

The heterogeneous relationship diagram of the network is introduced to represent network entities, configuration and the relationship among the network entities, so that structural description of the network system is realized. The first state information of the heterogeneous relational graph at the current moment is processed through the network performance prediction model, hidden state information with global state information can be obtained, and more accurate network performance indexes can be obtained based on the hidden state information with the global state information. The first state information of the heterogeneous relational graph of the target network at the current moment at the next moment is predicted based on the hidden state information, the performance index of the target network at the subsequent moment and the first state information of the heterogeneous relational graph of the current moment at the subsequent moment can be predicted based on the first state information of the heterogeneous relational graph of the current moment at the next moment according to a recursive mode, the time sequence expression of the performance index of the target network is achieved, and the network performance prediction model of the target network has systematic expression capability.

In a possible embodiment, the network performance prediction model of the target network includes an encoding network, an information delivery network, and a decoding network, and the second state information of the current heterogeneous relationship diagram is obtained by processing the current heterogeneous relationship diagram and the first state information of the current heterogeneous relationship diagram through the network performance prediction model of the target network, including:

coding the first state information of the heterogeneous relationship diagram at the current moment through a coding network to obtain third state information of the heterogeneous relationship diagram at the current moment; performing multiple rounds of information transmission operation on the heterogeneous relationship diagram at the current moment, the third state information of the heterogeneous relationship diagram at the current moment and the hidden state information of the heterogeneous relationship diagram at the previous moment through an information transmission network to obtain the hidden state information of the heterogeneous relationship diagram at the current moment; and decoding the hidden state information of the heterogeneous relationship diagram at the current moment through a decoding network to obtain second state information of the heterogeneous relationship diagram at the current moment.

It should be understood that the heterogeneous relationship diagram at the current time includes a plurality of nodes and side information between the plurality of nodes, and the encoding operation is performed on the first state information of the heterogeneous relationship diagram at the current time through the encoding network to obtain the third state information of the heterogeneous relationship diagram at the current time, specifically, the encoding operation is performed on the first state information of each node in the heterogeneous relationship diagram at the current time.

Since the dimensions of the first state information of different nodes in the heterogeneous relationship diagram at the current time may be different, in order to facilitate subsequent processing, the first state information of each node in the heterogeneous relationship diagram at the current time is encoded by the encoding network to obtain the third state information of each node, and the dimensions of the third state information of different nodes are the same. And performing multi-round information transmission through the information transmission network, so that the hidden state information of each node comprises the state information of other nodes, and further more accurate performance indexes can be obtained based on the hidden state information.

In one possible embodiment, the method for obtaining hidden state information of a heterogeneous relationship diagram at a current time by performing multiple rounds of information transfer operations on the heterogeneous relationship diagram at the current time, third state information of the heterogeneous relationship diagram at the current time and hidden state information of the heterogeneous relationship diagram at a previous time through an information transfer network includes:

obtaining fourth state information of the heterogeneous relational graph at the current moment based on the third state information of the heterogeneous relational graph at the current moment and the hidden state information of the heterogeneous relational graph at the previous moment; the fourth state information of the heterogeneous relationship diagram at the current moment comprises the fourth state information of each node in the heterogeneous relationship diagram at the current moment; when the ith round of information transmission is carried out, state information S of neighbor nodes of each node in the heterogeneous relation graph at the current moment is acquired based on side information among a plurality of nodes _i (ii) a i is an integer greater than 0; state information S of neighbor nodes based on each node _i And state information S of each node _i Obtaining state information S of each node _i+1 (ii) a When i is 1, the state information S of each node in the heterogeneous relationship graph at the current moment _i Fourth state information for each node; when the ith round of information transfer is the last round of information transfer in the multi-round of information transfer, the state information S of each node _i+1 And for the hidden state information of each node, the hidden state information of the heterogeneous relationship graph at the current moment comprises the hidden state information of each node in the heterogeneous relationship graph at the current moment.

In order to obtain a more accurate performance index, before information transmission is performed, hidden state information at a previous moment needs to be introduced, the hidden state information at the previous moment and third state information of a heterogeneous relationship diagram at the current moment are integrated together to obtain fourth state information of the heterogeneous relationship diagram at the current moment, and then information transmission is performed based on the fourth state information of the heterogeneous relationship diagram at the current moment.

In one possible embodiment, the state information S is based on the neighbor nodes of each node _i And state information S of each node _i Obtaining state information S of each node _i+1 The method comprises the following steps:

state information S of neighbor nodes to each node _i Performing aggregation operation to obtain aggregated state information of each node; converting the aggregated state information of each node to obtain converted state information of each node; state information S for each node based on converted state information of each node _i Updating to obtain the state information S of each node _i+1 。

Since the aggregate information of each node is the state information S of the neighbor nodes for each node _i Meaning expressed by aggregated information of each node and status information S of the node _i The expressed meanings are not consistent, so that the aggregated information needs to be converted so that the meaning expressed by the converted state information is the same as the state information S of the node _i The meanings expressed are consistent. By pairing nodes based on converted state informationState information S of _i Updating to obtain status information S _i+1 Including state information of its neighbor nodes. After multiple rounds of information transfer, information with a global state is available for each node.

In one possible embodiment, the determining the first state information of the heterogeneous relationship diagram at the current time at the next time based on the second state information and the first state information of the heterogeneous relationship diagram at the current time further includes:

determining dynamic information variation from second state information of the heterogeneous relationship diagram at the current moment; and obtaining first state information of the heterogeneous relational graph at the current moment at the next moment according to the dynamic information variation, the dynamic information at the current moment and the static information.

After the first state information of the heterogeneous relationship diagram at the current time at the next time is obtained, according to a recursive manner, the performance index of the target network at the next time and the first state information of the heterogeneous relationship diagram at the current time at the next time, even the performance index of the target network at each subsequent time and the first state information of the heterogeneous relationship diagram at each subsequent time, can be predicted according to the first state information of the heterogeneous relationship diagram at the current time at the next time.

In a second aspect, an embodiment of the present application provides a method for training a network performance prediction model, including:

acquiring a plurality of heterogeneous relationship graphs and fifth state information of each heterogeneous relationship graph in the plurality of heterogeneous relationship graphs, wherein the plurality of heterogeneous relationship graphs are acquired at different moments aiming at the same network, and each heterogeneous relationship graph is used for representing a plurality of nodes participating in network operation in the moment of acquiring each heterogeneous relationship graph and relationships among the plurality of nodes; and performing multiple rounds of training on the initial prediction model based on the multiple heterogeneous relationship graphs and the fifth state information of each heterogeneous relationship graph in the multiple heterogeneous relationship graphs to obtain a network performance prediction model.

In one possible embodiment, performing multiple rounds of training on the initial prediction model based on the multiple heterogeneous relationship graphs and the fifth state information of each of the multiple heterogeneous relationship graphs to obtain a network performance prediction model, includes:

dividing a plurality of heterogeneous relationship graphs into a plurality of groups, wherein each group comprises at least one heterogeneous relationship graph; during the p-th round of training, the fifth state information and the hidden state information H 'of the heterogeneous relational graph of the p-th group in the plurality of groups' _p Input to a Performance prediction model M _p The processing is carried out to obtain at least one sixth state information corresponding to the p-th group of heterogeneous relational graph, the sixth state information comprises a performance index and a dynamic information variation, and p is an integer larger than 0; inputting the performance index and the dynamic information variation included in the at least one sixth state information into a loss function for calculation to obtain a loss value L _p (ii) a Based on the loss value L _p Adjusting performance prediction model M _p To obtain a performance prediction model M _p+1 (ii) a When p is 1, the performance prediction model M _p Hidden state information H 'for the initial prediction model' _p Is an empty set; when p is greater than 1, hiding the state information H' _p Is obtained according to the last heterogeneous relation graph in the p-1 group, and when the p-th training is the last training, the performance prediction model M _p+1 Is a network performance prediction model.

In one possible embodiment, the initial prediction model comprises an initial coding network, an initial information transfer network and an initial decoding network, and the performance prediction model M _p Comprising a coding network M _1p Information transfer network M _2p And decoding network M _3p (ii) a Each heterogeneous relationship graph comprises a plurality of nodes and side information among the nodes; the fifth state information of each heterogeneous relationship diagram comprises fifth state information of each node in the heterogeneous relationship diagram, and the fifth state information and hidden state information H 'of the heterogeneous relationship diagram of the p-th group in the multiple groups' _p Input to a Performance prediction model M _p To obtain at least one corresponding heterogeneous relation graph of the p-th groupSixth state information including:

performing the following processing on the fifth state information of the jth heterogeneous relationship diagram in the pth group to obtain sixth state information corresponding to the jth heterogeneous relationship diagram:

by coding the network M _1p Processing the fifth state information of each node in the jth heterogeneous relationship graph to obtain seventh state information of the jth heterogeneous relationship graph; passing through an information transfer network M based on side information between a plurality of nodes included in a jth heterogeneous relationship graph _2p For the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation to obtain hidden state information H of the jth heterogeneous relational graph _j+1 (ii) a By decoding the network M _3p Hidden state information H for jth heterogeneous relational graph _j+1 Decoding to obtain sixth state information corresponding to the jth heterogeneous relation graph; when j is 1, hidden state information H of j-1 th heterogeneous relation graph _j Is hidden state information H' _p (ii) a When p is 1, the network M is coded _1p For the initial coding of the network, the information transfer network M _2p For the initial information transfer network, decoding the network M _3p The network is initially decoded.

In one possible embodiment, the information delivery network M is based on side information between a plurality of nodes included in the jth heterogeneous relationship graph _2p For the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation to obtain hidden state information H of the jth heterogeneous relational graph _j+1 The method comprises the following steps:

seventh state information based on jth heterogeneous relational graph and hidden state information H of jth-1 heterogeneous relational graph _j Obtaining eighth state information of a jth heterogeneous relation graph; the eighth state information of the jth heterogeneous relationship graph comprises eighth state information of each node in the jth heterogeneous relationship graph; during the k round of information transmission, acquiring the neighbor nodes of each node in the j heterogeneous relationship graph based on the side information among a plurality of nodes included in the j heterogeneous relationship graphStatus information S _k (ii) a The neighbor node of each node is a node with an edge between the neighbor node and the node; k is an integer greater than 0; state information S of neighbor nodes based on each node in jth heterogeneous relation graph _k And state information S of each node _k Obtaining state information S of each node _k+1 (ii) a When k is 1, state information S of the neighbor node _k As eighth state information of the neighboring node, state information S of each node _k Eighth state information for each node; when the kth round of information transfer is the last round of information transfer in the multi-round of information transfer, the state information S of each node _i+1 Hidden state information of jth heterogeneous relational graph H as hidden state information of each node _j+1 Including hidden state information for each node in the jth heterogeneous relationship graph.

In one possible embodiment, the state information S of the neighbor nodes of each node in the jth heterogeneous relationship graph is based on _k And state information S of each node _k Obtaining state information S of each node _k+1 The method comprises the following steps:

state information S of neighbor nodes to each node _k Performing aggregation operation to obtain aggregated state information of each node; converting the aggregated state information of each node to obtain converted state information of each node; updating the state information of each node based on the converted state information of each node to obtain the state information S of each node _k+1 。

In one possible embodiment, the plurality of nodes of the network comprises: network devices, links, queues, and paths. In some scenarios, the plurality of nodes of the target network also include a flow. The network device may include a switch and a server.

It should be understood that the network performance prediction model of the first aspect may be trained by using the method of the second aspect, and the training process is consistent with the process of predicting the performance index by using the network performance prediction model, so that the beneficial effects of the content of the second aspect can be referred to the beneficial effects of the content of the first aspect, and will not be described herein.

In a third aspect, an embodiment of the present application provides a prediction apparatus, including:

the system comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a heterogeneous relationship graph at the current time and first state information of the heterogeneous relationship graph at the current time aiming at a target network, and the heterogeneous relationship graph at the current time is used for representing a plurality of nodes participating in network operation in the target network at the current time and relations among the nodes;

the prediction unit is used for processing the heterogeneous relationship diagram at the current moment and the first state information of the heterogeneous relationship diagram at the current moment through a network performance prediction model of the target network to obtain second state information of the heterogeneous relationship diagram at the current moment, wherein the second state information comprises a performance index of the target network at the current moment;

and the determining unit is used for determining the first state information of the heterogeneous relationship diagram at the current moment at the next moment based on the second state information and the first state information of the heterogeneous relationship diagram at the current moment.

In a possible embodiment, the network performance prediction model of the target network includes an encoding network, an information delivery network, and a decoding network, and the prediction unit is specifically configured to:

In a possible embodiment, the heterogeneous relationship graph at the current time includes a plurality of nodes of the target network and side information between the plurality of nodes, where the side information between the plurality of nodes is used to represent a relationship between the plurality of nodes, and in an aspect that the heterogeneous relationship graph at the current time, the third state information of the heterogeneous relationship graph at the current time, and the hidden state information of the heterogeneous relationship graph at the previous time are subjected to multiple rounds of information transfer operations through the information transfer network to obtain the hidden state information of the heterogeneous relationship graph at the current time, the prediction unit is specifically configured to:

obtaining fourth state information of the heterogeneous relational graph at the current moment based on the third state information of the heterogeneous relational graph at the current moment and the hidden state information of the heterogeneous relational graph at the previous moment; the fourth state information of the heterogeneous relationship diagram at the current moment comprises the fourth state information of each node in the heterogeneous relationship diagram at the current moment; when the ith round of information transmission is carried out, state information S of neighbor nodes of each node in the heterogeneous relation graph at the current moment is acquired based on side information among a plurality of nodes _i (ii) a i is an integer greater than 0; state information S of neighbor nodes based on each node _i And state information S of each node _i Obtaining state information S of each node _i+1 (ii) a When i is 1, the state information S of each node in the heterogeneous relationship graph at the current moment _i Fourth state information for each node; when the ith round of information transfer is the last round of information transfer in the multiple rounds of information transfer, the state information S of each node _i+1 And for the hidden state information of each node, the hidden state information of the heterogeneous relationship graph at the current moment comprises the hidden state information of each node in the heterogeneous relationship graph at the current moment.

In one possible embodiment, the state information S is based on the state information S of each node' S neighbor nodes _i And state information S of each node _i Obtaining state information S of each node _i+1 In an aspect, the prediction unit is specifically configured to:

In one possible embodiment, the first state information of the heterogeneous relationship graph at the current time includes static information and dynamic information at the current time, the second state information further includes a state information variation, and the dynamic information variation is a variation between the dynamic information at the next time and the dynamic information at the current time, and the determining unit is specifically configured to include:

In a fourth aspect, an embodiment of the present application provides an exercise device, including:

the acquiring unit is used for acquiring a plurality of heterogeneous relationship graphs and fifth state information of each heterogeneous relationship graph in the plurality of heterogeneous relationship graphs, the plurality of heterogeneous relationship graphs are acquired at different moments aiming at the same network, and each heterogeneous relationship graph is used for representing a plurality of nodes participating in network operation in the moment of acquiring each heterogeneous relationship graph and relationships among the plurality of nodes;

and the training unit is used for carrying out multi-round training on the initial prediction model based on the plurality of heterogeneous relationship graphs and the fifth state information of each heterogeneous relationship graph in the plurality of heterogeneous relationship graphs so as to obtain a network performance prediction model.

In a possible embodiment, the training unit is specifically configured to:

dividing a plurality of heterogeneous relationship graphs into a plurality of groups, wherein each group comprises at least one heterogeneous relationship graph; during the p-th round of training, the fifth state information and the hidden state information H 'of the heterogeneous relational graph of the p-th group in the plurality of groups' _p Input to a Performance prediction model M _p To obtain at least one sixth state information corresponding to the p-th group of heterogeneous relationship graphs, and the sixth stateThe information comprises a performance index and dynamic information variation, and p is an integer greater than 0; inputting the performance index and the dynamic information variation included in the at least one sixth state information into a loss function for calculation to obtain a loss value L _p (ii) a Based on the loss value L _p Adjusting performance prediction model M _p To obtain a performance prediction model M _p+1 (ii) a When p is 1, the performance prediction model M _p Hidden state information H 'for the initial prediction model' _p Is an empty set; when p is greater than 1, hiding the state information H' _p Is obtained according to the last heterogeneous relation graph in the p-1 group, and when the p-th training is the last training, the performance prediction model M _p+1 Is a network performance prediction model.

In one possible embodiment, the initial prediction model comprises an initial coding network, an initial information transfer network and an initial decoding network, and the performance prediction model M _p Comprising a coding network M _1p Information transfer network M _2p And decoding network M _3p (ii) a Each heterogeneous relationship graph comprises a plurality of nodes and side information among the nodes; the fifth state information of each heterogeneous relationship graph comprises fifth state information of each node in the heterogeneous relationship graph, and the fifth state information and hidden state information H 'of the heterogeneous relationship graph of the p-th group in the plurality of groups' _p Input to a Performance prediction model M _p In order to obtain at least one sixth state information corresponding to the p-th group of heterogeneous relationship graphs, the training unit is specifically configured to:

by coding the network M _1p Processing the fifth state information of each node in the jth heterogeneous relationship graph to obtain seventh state information of the jth heterogeneous relationship graph; passing through an information transfer network M based on side information between a plurality of nodes included in a jth heterogeneous relationship graph _2p For the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation toObtaining the hidden state information H of the jth heterogeneous relational graph _j+1 (ii) a By decoding the network M _3p Hidden state information H for jth heterogeneous relational graph _j+1 Decoding to obtain sixth state information corresponding to the jth heterogeneous relation graph; when j is 1, hidden state information H of j-1 th heterogeneous relation graph _j Is hidden state information H' _p (ii) a When p is 1, the network M is coded _1p For the initial coding of the network, the information transfer network M _2p For the initial information transfer network, decoding the network M _3p The network is initially decoded.

In one possible embodiment, the information delivery network M is based on side information between a plurality of nodes comprised by the jth heterogeneous relational graph _2p For the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation to obtain hidden state information H of the jth heterogeneous relational graph _j+1 In an aspect of (a), the training unit is specifically configured to:

seventh state information based on jth heterogeneous relationship graph and hidden state information H of jth heterogeneous relationship graph-1 _j Obtaining eighth state information of a jth heterogeneous relation graph; the eighth state information of the jth heterogeneous relationship graph comprises eighth state information of each node in the jth heterogeneous relationship graph; when the kth round of information transmission is carried out, the state information S of the neighbor node of each node in the jth heterogeneous relation graph is obtained based on the side information among a plurality of nodes included in the jth heterogeneous relation graph _k (ii) a The neighbor node of each node is a node with an edge between the neighbor node and the node; k is an integer greater than 0; state information S of neighbor nodes based on each node in jth heterogeneous relation graph _k And state information S of each node _k Obtaining state information S of each node _k+1 (ii) a When k is 1, the state information S of the neighbor node _k As eighth state information of the neighboring node, state information S of each node _k Eighth state information for each node; when the kth round of information transfer is the last round of information transfer in the multi-round of information transfer, the state information S of each node _i+1 For the hidden state information of each node,hidden state information H of jth heterogeneous relational graph _j+1 Including hidden state information for each node in the jth heterogeneous relationship graph.

In one possible embodiment, the state information S of the neighbor nodes based on each node in the jth heterogeneous relationship graph _k And state information S of each node _k Obtaining state information S of each node _k+1 In an aspect of (a), the training unit is specifically configured to:

In a fifth aspect, an embodiment of the present application further provides an electronic device, including a processor and a memory, where the processor is connected to the memory, where the memory is used to store a program code, and the processor is used to call the program code to execute part or all of the method of the first aspect or the second aspect.

In a sixth aspect, an embodiment of the present application further provides a chip system, where the chip system is applied to an electronic device; the chip system comprises one or more interface circuits, and one or more processors; the interface circuit and the processor are interconnected through a line; the interface circuit is to receive a signal from a memory of the electronic device and to send the signal to the processor, the signal comprising computer instructions stored in the memory; when the processor executes the computer instructions, the electronic device performs part or all of the method of the first or second aspect.

In a seventh aspect, this application embodiment further provides a computer-readable storage medium, where a computer program is stored, and the computer program is executed by a processor to implement part or all of the method in the first aspect or the second aspect.

In an eighth aspect, the present application further provides a computer program executed to implement part or all of the method described in the first aspect or the second aspect.

These and other aspects of the present application will be more readily apparent from the following description of the embodiments.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

FIG. 1a is a schematic diagram of a system architecture according to an embodiment of the present application;

FIG. 1b is a schematic diagram of another system architecture according to an embodiment of the present application;

fig. 2 is a schematic flowchart of a network performance prediction method according to an embodiment of the present application;

fig. 3a is a schematic diagram of a heterogeneous relationship diagram provided in an embodiment of the present application;

FIG. 3b is a schematic diagram of another isomeric relationship diagram provided in an embodiment of the present application;

fig. 4 is a schematic flowchart illustrating a specific process of a step in a network performance prediction method according to an embodiment of the present application;

FIG. 5 is a schematic diagram of another isomeric relationship diagram provided in an embodiment of the present application;

FIG. 6 is a schematic diagram of a process for using a network performance prediction model according to an embodiment of the present application;

fig. 7 is a schematic diagram of a data center network according to an embodiment of the present application;

fig. 8 is a schematic diagram of a heterogeneous relationship diagram of a data center network according to an embodiment of the present application;

fig. 9 is a schematic flowchart of a network performance prediction model training method according to an embodiment of the present disclosure;

fig. 10 is a schematic structural diagram of a prediction apparatus according to an embodiment of the present application;

FIG. 11 is a schematic structural diagram of an exercise device according to an embodiment of the present disclosure;

fig. 12 is a schematic structural diagram of an electronic device according to an embodiment of the present application;

fig. 13 is a schematic structural diagram of another electronic device according to an embodiment of the present application.

Detailed Description

The terms "first," "second," "third," and "fourth," etc. in the description and claims of this application and in the accompanying drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "include" and "have," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements listed, but may alternatively include other steps or elements not listed, or inherent to such process, method, article, or apparatus.

Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.

"plurality" means two or more. "and/or" describes the association relationship of the associated objects, meaning that there may be three relationships, e.g., a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship.

Embodiments of the present application are described below with reference to the drawings.

Referring to fig. 1a, the present embodiment provides a system architecture 100. The data collection device 160 is configured to collect a plurality of heterogeneous relationship graphs and a plurality of state information corresponding to the plurality of heterogeneous relationship graphs, and store the state information in the database 130. The heterogeneous relationship graphs are determined at different moments for the same network, and each heterogeneous relationship graph in the heterogeneous relationship graphs is used for representing a plurality of nodes participating in network operation at the moment corresponding to the heterogeneous relationship graph and relationships among the nodes. Each heterogeneous relationship graph comprises a plurality of nodes and side information among the nodes, and the side information among the nodes is used for representing the relationship among the nodes. Training device 120 generates network performance prediction model 101 based on the plurality of heterogeneous relational graphs and state information for the plurality of heterogeneous relational graphs maintained in database 130. How the training device 120 obtains the network performance prediction model 101 based on the plurality of heterogeneous relationship graphs and the state information thereof can refer to the relevant description of the subsequent embodiments, and will not be described here. The network performance prediction model 101 is capable of predicting a performance index of the target network at the current time based on the heterogeneous relationship diagram of the current time input into the target network and the first state information thereof. Further, the network performance prediction model 101 may also predict the state information of the heterogeneous relationship diagram at the next time based on the heterogeneous relationship diagram at the current time of the input target network and the first state information thereof.

The network performance prediction model 101 obtained by the training device 120 may be applied to different systems or devices. In fig. 1a, the execution device 110 is configured with an I/O interface 112 to interact with external devices, and a "user" can input data to the I/O interface 112 through the client device 140, wherein the data includes a current-time heterogeneous relationship diagram of a target network and state information thereof.

The execution device 110 may call data, code, etc. from the data storage system 150 and may store data, instructions, etc. in the data storage system 150.

The computing module 111 processes the input data by using the network performance prediction model 101, and the processing process specifically includes:

coding the first state information of the heterogeneous relational graph at the current moment to obtain third state information of the heterogeneous relational graph at the current moment; performing multiple rounds of information transmission operation based on the third state information of the heterogeneous relationship diagram at the current moment and the hidden state information of the heterogeneous relationship diagram at the previous moment to obtain the hidden state information of the heterogeneous relationship diagram at the current moment; decoding the hidden state information of the heterogeneous relationship diagram at the current moment to obtain second state information of the heterogeneous relationship diagram at the current moment, wherein the second state information of the heterogeneous relationship diagram at the current moment comprises a performance index of the target network at the current moment, and determining first state information of the heterogeneous relationship diagram at the current moment at the next moment based on the second state information of the heterogeneous relationship diagram at the current moment and the first state information of the heterogeneous relationship diagram at the current moment.

Finally, the I/O interface 112 returns the processing results, such as the performance metrics of the target network at the current time, to the client device 140 for presentation to the user.

Further, the training device 120 may generate corresponding network performance prediction models 101 for different networks based on heterogeneous relationship maps of the different networks and state information thereof, so as to provide better results for users.

In the case shown in FIG. 1a, the user may manually specify the current-time heterogeneous relationship diagram of the target network in the input execution device 110 and its state information, e.g., operating in an interface provided by the I/O interface 112. In another case, the client device 140 may automatically input the current heterogeneous relationship diagram of the target network and the status information thereof to the I/O interface 112, and obtain the performance index of the current target network and the status information of the next heterogeneous relationship diagram, and if the client device 140 automatically inputs data and needs to obtain the authorization of the user, the user may set a corresponding right in the client device 140. The user may view, at the client device 140, the result output by the execution device 110, such as the performance index of the target network at the current time and the state information of the heterogeneous relationship diagram at the next time, where the specific presentation form may be a specific manner such as display, sound, and action. The client device 140 may also serve as a data acquisition end to store the acquired heterogeneous relationship diagram of the target network at the current time and the state information thereof as training data in the database 130.

It should be noted that fig. 1a is only a schematic diagram of a system architecture provided by an embodiment of the present invention, and the position relationship between the devices, modules, etc. shown in the diagram does not constitute any limitation, for example, in fig. 1a, the data storage system 150 is an external memory with respect to the execution device 110, and in other cases, the data storage system 150 may also be disposed in the execution device 110.

Referring to fig. 1b, fig. 1b is a schematic diagram of a system architecture provided in the embodiment of the present application. As shown in fig. 1b, the system architecture includes a prediction device 10 and a computing device 20,

the prediction apparatus 10 may be a terminal device, such as a smart phone, a desktop computer, a notebook computer, a tablet computer, or the like. The computing device 20 may be a server, a cluster of servers, a cloud server, or other form of equipment having computing capabilities.

The prediction apparatus 10 transmits a model acquisition request to the calculation apparatus 20. In one example, the model acquisition request includes an identification of the target network, and the computing device 20 stores a plurality of network performance prediction models corresponding to the plurality of networks. The computing device 20 obtains a network performance prediction model corresponding to the target network from the plurality of network performance prediction models according to the identifier of the target network. In another embodiment, the model obtaining request carries training data, where the training data includes a plurality of different-time heterogeneous relationship graphs and state information thereof obtained for the target network, and the computing device 20 obtains the network performance prediction model corresponding to the target network based on the plurality of different-time heterogeneous relationship graphs and the state information thereof. The computing device 20 sends a response message carrying the network performance prediction model corresponding to the target network to the prediction device 10. The prediction apparatus 10 predicts the performance index of the target network at the current time and the first state information of the heterogeneous relationship diagram at the current time based on the heterogeneous relationship diagram at the current time of the target network and the first state information thereof based on a network performance prediction model corresponding to the target network.

In some embodiments, the prediction apparatus 10 has a strong computing capability, and the prediction apparatus 10 obtains a network performance prediction model corresponding to the target network based on the heterogeneous relationship diagram at a plurality of different time instants and the state information thereof. The prediction apparatus 10 predicts the performance index of the target network at the current time and the state information of the heterogeneous relationship diagram at the next time based on the heterogeneous relationship diagram at the current time of the target network and the first state information thereof based on a network performance prediction model corresponding to the target network.

For predicting the performance index of the target network at the current time and the first state information of the heterogeneous relationship diagram at the current time at the next time based on the network performance prediction model corresponding to the target network and based on the heterogeneous relationship diagram at the current time of the target network and the first state information of the heterogeneous relationship diagram at the current time, the method specifically comprises the following steps:

encoding the first state information of the heterogeneous relationship diagram at the current moment to obtain third state information of the heterogeneous relationship diagram at the current moment; performing multiple rounds of information transmission operation based on the third state information of the heterogeneous relationship diagram at the current moment and the hidden state information of the heterogeneous relationship diagram at the previous moment to obtain the hidden state information of the heterogeneous relationship diagram at the current moment; decoding the hidden state information of the heterogeneous relationship diagram at the current moment to obtain second state information of the heterogeneous relationship diagram at the current moment, wherein the second state information of the heterogeneous relationship diagram at the current moment comprises a performance index of the target network at the current moment, and determining first state information of the heterogeneous relationship diagram at the current moment at the next moment based on the second state information of the heterogeneous relationship diagram at the current moment and the first state information of the heterogeneous relationship diagram at the current moment.

It can be seen that by introducing a heterogeneous relationship diagram of a network, the heterogeneous relationship diagram is used for characterizing network entities, configurations and relationships among the network entities, so that structural description of a network system is realized. The state information of the heterogeneous relational graph is transmitted based on the heterogeneous relational graph, so that the state information of each node in the heterogeneous relational graph comprises the state information of other nodes in the heterogeneous relational graph, namely the state information of each node in the heterogeneous relational graph is global state information, performance index prediction is carried out based on the global state information, and accurate performance indexes can be predicted. The state information of the next moment and the state information of the subsequent moments can be predicted based on the state information of the current moment, and further the performance index of the target network of the next moment and the performance index of the target network of the subsequent moments can be predicted.

Referring to fig. 2, fig. 2 is a schematic flowchart of a network performance prediction method according to an embodiment of the present disclosure. As shown in fig. 2, the method includes:

s201, the prediction device obtains a current heterogeneous relationship diagram and first state information of the current heterogeneous relationship diagram aiming at a target network, wherein the current heterogeneous relationship diagram is used for representing the relationship among a plurality of nodes participating in network operation in the current target network.

Alternatively, the prediction means may be the execution device 110 in fig. 1a, or the prediction means 10 in fig. 1 b.

It is noted that the nodes in the heterogeneous relationship graph include physical nodes and virtual nodes. Where a physical node represents a network node having configuration information including, but not limited to, switches with buffers, servers, links (ports), queues. Virtual nodes represent nodes that are related to network performance including, but not limited to, flows, paths.

The state information is not the same for different nodes. State information such as switches include, but are not limited to: cache size, cache management policies, etc. The status information of the link (port) includes but is not limited to: bandwidth, propagation delay, queue scheduling policy, cache management control factors, and the like. The queue's state information includes, but is not limited to: scheduling weights, Active Queue Management (AQM) policies and their parameters, etc. The state information of the path includes, but is not limited to: aggregate traffic size, hop count, delay, throughput, jitter, packet loss, etc. The state information of the flow includes, but is not limited to: start time, stream size, remaining stream size, stream lifetime, etc.

The nodes contained in the heterogeneous relationship graph at different time instants may not be the same. A heterogeneous relationship graph is used to characterize a plurality of nodes and relationships between the plurality of nodes participating in network operations in a network at a time. Since the nodes participating in the network operation in the target network may not be the same at different times and the relationships between the nodes participating in the network operation are also different, the heterogeneous relationship graphs at different times may not be the same.

The side information between the nodes is used to characterize the relationship between the nodes. Whether two nodes have a relationship between them is determined based on domain knowledge. If the two nodes have a relationship, the two nodes are adjacent nodes.

For example, assume that the target network includes switch S0 and switch S1, where switch S0 includes link L0 and link L1, and switch S1 includes link L2 and link L3. Link L0 is configured with queue Q0 and queue Q1, link L1 is configured with queue Q2 and queue Q3, link L2 is configured with queue Q4 and queue Q5, and link L3 is configured with queue Q6 and queue Q7.

At a first time, path P0 passes through queue Q0 and queue Q4, and flows F0 and F1 are transmitted through path P0. Path P1 passes through queue Q1 and queue Q7, and flow F2 is transmitted through path P1. Fig. 3a illustrates a heterogeneous relationship diagram at a first time. As shown in fig. 3a, node S0 has a relationship with link L0, link L0 has a relationship with both queue Q0 and queue Q1, node S1 has a relationship with link L2, link L2 has a relationship with queue Q4, link L3 has a relationship with queue Q7, path P0 has a relationship with queue Q0 and queue Q4, path P1 has a relationship with queue Q1 and queue Q7, path P0 has a relationship with flow F0 and flow F1, and path P1 has a relationship with flow F2. The dashed lines in fig. 3a indicate that the corresponding node is not participating in the operation of the network at the first moment.

At a second time, path P0 passes through queue Q0, queue Q1, and queue Q4, and flow F0 is transmitted through path P0. Path P1 passes through queue Q2, queue Q3, and queue Q7, and flow F2 and flow F3 are transmitted through path P1. Fig. 3b illustrates a heterogeneous relationship diagram at a second time. As shown in FIG. 3b, node S0 has a relationship with link L0 and link L1, link L0 has a relationship with both queue Q0 and queue Q1, and link L1 has a relationship with both queue Q2 and queue Q3. Node S1 has a relationship with link L2 and link L3, link L2 has a relationship with queue Q4, link L3 has a relationship with queue Q7, path P0 has a relationship with queue Q0, queue Q1 and queue Q4, path P1 has a relationship with queue Q2, queue Q3 and queue Q7, path P0 has a relationship with flow F0, and path P1 has a relationship with flow F2 and flow F3. The dashed lines in fig. 3b indicate that the corresponding node is not participating in the operation of the target network at the second moment in time.

As can be seen from fig. 3a and fig. 3b, the heterogeneous relationship diagrams at different time instants may be different for the same network.

S202, the prediction device processes the heterogeneous relationship diagram at the current moment and the first state information of the heterogeneous relationship diagram at the current moment through the network performance prediction model of the target network to obtain second state information of the heterogeneous relationship diagram at the current moment, wherein the second state information comprises a performance index of the target network at the current moment.

The network performance prediction model comprises an encoding network, an information transmission network and a decoding network. The specific process of S202 is shown in fig. 4, and includes:

s401, the prediction device performs coding operation on the first state information of the heterogeneous relationship diagram at the current time through a coding network to obtain third state information of the heterogeneous relationship diagram at the current time.

Since the heterogeneous relationship graph includes a plurality of different types of nodes, the state information is not the same for the different types of nodes. Accordingly, the dimensions of the state information may not be the same. In this application, the state information is represented by a vector. For different types of nodes, the dimensions of vectors corresponding to state information may be different, and for convenience of subsequent processing, the state information of each node in the heterogeneous relationship diagram needs to be processed, so that the dimensions of the state information processed by all nodes in the heterogeneous relationship diagram are the same. When the state information is represented by a vector, processing the vector corresponding to the state information of each node in the heterogeneous relational graph so as to ensure that the dimensions of the vectors corresponding to the state information of all nodes in the heterogeneous relational graph are the same. The specific treatment method comprises the following steps:

the prediction device processes the first state information of each node in the heterogeneous relationship graph at the current time by using the coding network to obtain the third state information of each node in the heterogeneous relationship graph at the current time. The third state information of the heterogeneous relationship diagram at the current time includes the third state information of each node in the heterogeneous relationship diagram at the current time.

Optionally, the coding network is implemented based on a neural network, such as a graph neural network.

S402, the prediction device carries out multiple rounds of information transmission operations on the heterogeneous relationship diagram at the current moment, the third state information of the heterogeneous relationship diagram at the current moment and the hidden state information of the heterogeneous relationship diagram at the previous moment through the information transmission network so as to obtain the hidden state information of the heterogeneous relationship diagram at the current moment.

It should be noted that, in order to obtain a more accurate performance index, before performing multiple rounds of information transfer operations, the prediction apparatus needs to process, for example, concatenate, the third state information of the heterogeneous relationship diagram at the current time and the hidden state information of the heterogeneous relationship diagram at the previous time to obtain the fourth state information of the heterogeneous relationship diagram at the current time.

Specific ways of splicing the third state information of the heterogeneous relationship diagram at the current time and the hidden state information of the heterogeneous relationship diagram at the previous time include, but are not limited to:

and for any node X in the heterogeneous relationship graph at the current moment, if the heterogeneous relationship graph at the previous moment has the node X, obtaining fourth state information of the node X according to the third state information and the hidden state information of the node X. For example, if the hidden state information of the node X is represented by a 3-dimensional vector and the third state information of the node X is represented by a 4-dimensional vector, the fourth state information of the node X is represented by a 7-dimensional vector obtained by the 3-dimensional vector representing the hidden state information of the node X and the 4-dimensional vector representing the third state information of the node X. And if the heterogeneous relational graph at the previous moment does not have the node X, determining the third state information of the node X as the fourth state information of the node X.

According to the method, the fourth state information of each node in the heterogeneous relationship graph at the current moment can be obtained. The fourth state information of the heterogeneous relationship diagram of the current state includes the third state information of each node in the heterogeneous relationship diagram at the current moment.

After the fourth state information of the heterogeneous relationship diagram at the current time is obtained, since the state information of each node only contains the state information related to the node and does not contain the state information of other nodes in the heterogeneous relationship diagram, in order to improve the accuracy of the predicted performance index, the prediction device performs multiple rounds of information transmission operations on the fourth state information of the heterogeneous relationship diagram at the current time based on the heterogeneous relationship diagram at the current time, so that the state information of each node of the heterogeneous relationship diagram at the current time contains the state information of other nodes. The heterogeneous relationship graph at the current moment comprises a plurality of nodes of the target network and side information among the nodes, and the side information among the nodes is used for representing the relationship among the nodes. The concrete method is as follows:

during the ith round of information, acquiring the state information S of the neighbor node of each node in the heterogeneous relationship graph based on the side information among a plurality of nodes included in the heterogeneous relationship graph at the current moment _i (ii) a The neighbor node of each node is a node with an edge between the neighbor node and the node; state information S of neighbor nodes based on each node _i And state information S of each node _i Obtaining state information S of each node _i+1 (ii) a i is an integer greater than 0;

when i is 1, each node S in the heterogeneous relationship graph at the current moment _i Fourth state information for each node; when the ith round of information transfer is the last round of information transfer in the multi-round of information transfer, the state information S of each node _i+1 The hidden state information of the heterogeneous relationship graph at the current time includes the hidden state information of each node in the heterogeneous relationship graph at the current time.

Further, based on the state information S of the neighbor nodes of each node _i And state information S of each node _i Obtaining state information S of each node _i+1 The method comprises the following steps:

state information S of neighbor nodes to each node _i Carrying out the polymerizationOperation, which may be implemented by an aggregation function, where the aggregation function may implement a summation operation and a function of a Recurrent Neural Network (RNN) to obtain aggregated state information of each node; converting the aggregated state information of each node, which can be realized by a conversion function, to obtain the converted state information of each node; state information S for each node based on converted state information of each node _i Updating, e.g. by updating a function, to obtain per-node state information S _i+1 。

Since the aggregate information of each node is the state information S of the neighbor nodes for each node _i Meaning expressed by aggregated information of each node and status information S of the node _i The expressed meanings are not consistent, so that the aggregated information needs to be converted so that the meaning expressed by the converted state information is the same as the state information S of the node _i The meanings expressed are consistent. Through the conversion operation, the characteristic information of the neighbor node of each node can be mapped into the same hidden space corresponding to each node, and the modeling capability of the graph neural network cannot be limited when the characteristic information of the neighbor node is uniformly operated by using an update function in the follow-up process.

It is to be noted here that the information transfer network comprises 3 parts for implementing the functions of aggregation function, conversion function and update function, respectively. I.e. based on the state information S of the neighbour nodes of each node _i And state information S of each node _i Obtaining state information S of each node _i+1 Is realized by a neural network of the graph. It should be understood that the aggregation function, the conversion function, and the update function may be implemented based on three neural networks, respectively.

It should be understood that the heterogeneous relationship graph may include the same type of nodes, for example, a plurality of nodes of a corresponding switch in the heterogeneous relationship graph are the same type of nodes, and a plurality of nodes of a corresponding path are the same type of nodes. For a plurality of nodes of the same type, the aggregation function, the conversion function and the update function are used the same, that is, the neural network is used the same.

For example, assume that the heterogeneous relationship diagram at the current time includes 7 nodes as shown in fig. 5. Performing 3 rounds of information transfer operations on the third state information of the heterogeneous relationship diagram shown in fig. 5, specifically including:

when the 1 st information transmission operation is carried out, for the node A, the neighbor nodes of the node A comprise the node B and the node D, and the predicting device obtains the state information S of the node A according to the third state information of the node B, the node D and the node A ₁ State information S of node A ₁ Containing the third state information of nodes A, B and D. For node B, the neighbor nodes of node B include nodes A and C, and the predicting device obtains the state information S of node B according to the third state information of nodes B, C and A ₁ State information S of node B ₁ Containing the third state information of nodes A, B and C. For the node C, the neighbor nodes of the node C include the nodes B, D, E and G, the predicting device obtains the state information S of the node C according to the third state information of the nodes B, C, D, E and G ₁ State information S of node C ₁ Containing the third state information of nodes B, C, D, E and G. For node D, the neighbor nodes of node D include nodes A, C and G, and the predicting device obtains the state information S of node D according to the third state information of nodes A, C, D and G ₁ State information S of node D ₁ Containing the third state information of nodes A, C, D and G. For the node E, the neighbor nodes of the node E comprise nodes F and C, and the prediction device obtains the state information S of the node E according to the third state information of the node C, the node E and the node F ₁ State information S of node E ₁ Containing the third state information of nodes E, F and C. For the node F, the neighbor nodes of the node F comprise a node E, and the predicting device obtains the state information S of the node F according to the third state information of the node E and the node F ₁ State information S of node F ₁ Containing third state information for nodes E and F. For the node G, the neighbor nodes of the node G comprise nodes D and C, and the predicting device obtains the state information S of the node G according to the third state information of the node C, the node D and the node G ₁ State information S of node G ₁ Containing the third state information of nodes C, G and D.

When the 2 nd round information transmission operation is carried out, for the node A, the prediction device is based on the state information S of the node B, the node D and the node A ₁ Obtaining state information S of node A ₂ State information S of node A ₂ Status information S comprising nodes A, B and D ₁ I.e. the status information S of the node A ₂ Third state information of A, B, C, D and G. For node B, the prediction means is based on the state information S of node B, node C and node A ₁ Obtaining state information S of node B ₂ State information S of node B ₂ Status information S comprising nodes A, B and C ₁ I.e. state information S of the node B ₂ Third state information containing A, B, C, D, E and G. For node C, the prediction means is based on the state information S of nodes B, C, D, E and G ₁ Obtaining state information S of node C ₂ State information S of node C ₂ Status information S including nodes B, C, D, E and G ₁ I.e. the status information S of the node C ₂ Third state information including a, B, C, D, E, F, and G. For node D, the prediction means is based on the state information S of nodes A, C, D and G ₁ Obtaining state information S of node D ₂ State information S of node D ₂ Status information S including nodes A, C and G ₁ I.e. the status information S of the node D ₂ Third state information including a, B, D, E, and G. For node E, the prediction means is based on the state information S of node C, node E and node F ₁ Obtaining state information S of node E ₂ State information S of node E ₂ Status information S including node F and node C ₁ I.e. the state information S of the node E ₂ Third state information containing A, B, C, E and G. For the node F, the neighbor nodes of the node F comprise a node E, and the prediction device is used for predicting the state information S of the node E and the node F ₁ Obtaining state information S of node F ₂ State information S of node F ₂ Status information S comprising nodes E and F ₁ I.e. the state information S of the node F ₂ Containing the third state information of nodes C, E and F. For node G, the prediction means is based on the state information S of node C, node D and node G ₁ Obtaining state information S of node G ₂ State information S of node G ₂ Status information S including nodes C, D and G ₂ I.e. state information S of node G ₁ Containing the third state information of nodes A, B, C, D, E and G.

In the 3 rd round of information transmission operation, for the node A, the prediction device is based on the state information S of the node B, the node D and the node A ₂ Obtaining state information S of node A ₃ State information S of node A ₃ Status information S comprising nodes A, B and D ₂ . For node B, the prediction means is based on the state information S of node B, node C and node A ₂ Obtaining state information S of node B ₃ State information S of node B ₃ Status information S comprising nodes A, B and C ₂ . For node C, the prediction means is based on the state information S of nodes B, C, D, E and G ₂ Obtaining state information S of node C ₃ State information S of node C ₃ Status information S including nodes B, C, D, E and G ₂ . For node D, the prediction means is based on the state information S of nodes A, C, D and G ₂ Obtaining state information S of node D ₃ State information S of node D ₃ Status information S including nodes A, C and G ₂ . For node E, the prediction means is based on the state information S of node C, node E and node F ₂ Obtaining state information S of node E ₃ State information S of node E ₃ Status information S including node F and node C ₂ . For the node F, the neighbor nodes of the node F comprise a node E, and the prediction device is used for predicting the state information S of the node E and the node F ₂ Obtaining state information S of node F ₃ State information S of node F ₃ Status information S comprising nodes E and F ₂ . For the node G, the neighbor nodes of the node G comprise nodes D and G, and the prediction device is used for predicting the state information S of the node C, the node D and the node G ₂ Obtaining state information S of node G ₃ State information S of node G ₃ Status information S including nodes C, D and G ₂ . At the moment, the state information S of each node in the heterogeneous relationship graph at the current moment ₃ Containing third state information of other nodes.

S403, the prediction device decodes the hidden state information of the heterogeneous relationship diagram at the current time through a decoding network to obtain second state information of the heterogeneous relationship diagram at the current time.

And the second state information of the heterogeneous relationship diagram at the current moment comprises the performance index of the target network at the current moment.

Alternatively, the decoding network may be implemented by a neural network, such as a graph neural network.

S203, the prediction device determines the first state information of the heterogeneous relationship diagram at the current time at the next time based on the second state information and the first state information of the heterogeneous relationship diagram at the current time.

Specifically, the first state information of the heterogeneous relationship diagram at the current time includes static information and dynamic information at the current time, the second state information further includes a state information variation, the dynamic information variation is a variation between the dynamic information at the next time and the dynamic information at the current time, and the prediction apparatus determines the dynamic information variation from the second state information of the heterogeneous relationship diagram at the current time; and obtaining first state information of the heterogeneous relational graph at the current moment at the next moment according to the dynamic information variation, the dynamic information at the current moment and the static information.

After obtaining the first state information of the heterogeneous relationship diagram at the next time, the prediction apparatus may perform recursive processing on the first state information of the heterogeneous relationship diagram at the next time based on the manners of S202 and S203 to predict the performance index of the target network at the next time and the first state information of the heterogeneous relationship diagram at the next time, even the performance index of the target network at each subsequent time and the first state information of the heterogeneous relationship diagram at the current time at each subsequent time.

It should be understood that the static information is information that does not change with time in the target network, such as total buffer size of the switch, link bandwidth, propagation delay, number of path hops, start time of the stream, and so on. The dynamic information is information that changes with time in the target network, such as path traffic size, remaining flow size, flow lifetime, and the like. If the configuration in the network changes over time, it should also be considered as dynamic information, such as active queue management parameters, queue weights, etc.

It can be seen that in the solution of the present application, a heterogeneous relationship diagram of a network is introduced to characterize network entities, configurations and relationships among the network entities, thereby implementing a structural description of a network system. The information transmission is carried out on the state information of the heterogeneous relational graph based on the heterogeneous relational graph, so that the state information of each node in the heterogeneous relational graph comprises the state information of other nodes in the heterogeneous relational graph, namely the state information of each node in the heterogeneous relational graph is global state information, the performance index prediction is carried out based on the global state information, and accurate performance indexes can be predicted. The first state information of the heterogeneous relational graph at the current moment can be predicted based on the first state information of the heterogeneous relational graph at the current moment, the performance index of the target network at the subsequent moment and the first state information of the heterogeneous relational graph at the current moment can be predicted based on the first state information of the heterogeneous relational graph at the current moment in a recursive mode, and the time sequence expression of the performance index of the target network is achieved.

The following describes specifically predicting the performance index of the target network at the current time by using a network performance prediction model of the target network in combination with the accompanying drawings.

As shown in fig. 6, the network performance prediction model includes an encoding network, an information delivery network, and a decoding network. The state information of the current time refers to first state information of a heterogeneous relationship graph of the current time acquired by aiming at the target network. The heterogeneous relationship graph at the current moment comprises a plurality of nodes and side information among the nodes, and the first state information of the heterogeneous relationship graph at the current moment comprises the first state information of each node in the heterogeneous relationship graph. Because the dimensions of the first state information of different nodes in the heterogeneous relationship diagram may be inconsistent, in order to facilitate subsequent processing, the first state information of each node in the heterogeneous relationship diagram at the current moment is input into the coding network for processing, so as to obtain third state information of each node; and the dimensions of the third state information of all the nodes in the heterogeneous relationship graph at the current moment are the same.

The specific process of obtaining the fourth state information of the heterogeneous relationship diagram at the current time based on the hidden state information at the previous time and the third state information of the heterogeneous relationship diagram at the current time may refer to the relevant description of S402, which is not described herein. And performing N-round prediction on the fourth state information of the relation graph at the current moment through an information transfer model to obtain the hidden state information of the heterogeneous relation graph at the current moment, so that the hidden state information of each node in the heterogeneous relation graph at the current moment not only contains the first state information of the node, but also contains the first state information of other nodes in the heterogeneous relation graph at the current moment. The specific process can be referred to the related description of step S402, and will not be described here.

And decoding the hidden state information of the heterogeneous relationship diagram at the current moment through a decoding network to obtain second state information of the heterogeneous relationship diagram at the current moment. The first state information of the heterogeneous relationship diagram at the current time comprises the performance index and the dynamic information variation of the target network at the current time. The first state information of the heterogeneous relationship diagram at the current moment comprises static information and dynamic information of the heterogeneous relationship diagram at the current moment, and the dynamic information variation is the variation of the dynamic information of the heterogeneous relationship diagram at the current moment relative to the dynamic information of the heterogeneous relationship diagram at the current moment. The dynamic information of the heterogeneous relationship diagram at the current moment at the next moment can be obtained based on the dynamic information and the dynamic information variation of the heterogeneous relationship diagram at the current moment, and the first state information of the heterogeneous relationship diagram at the current moment at the next moment can be obtained based on the static information of the heterogeneous relationship diagram at the current moment and the dynamic information of the heterogeneous relationship diagram at the current moment.

In one particular example, fig. 7 is a schematic diagram of a data center network having a leaf-spine topology. As shown in fig. 7, the target network includes switch S0, switch S1, switch S2, switch S3, switch S4, switch S5, server H0, server H1, server H2, server H3, server H4, server H5, server H6, and server H7. The cache of the switch is shared among all ports, that is, the cache is shared by all ports of the switch, a queue of each port is managed by a scheduling policy, a stream is sent from a source server to a destination server, a Data Center Transmission Control Protocol (DCTCP) is run on the server to perform congestion control, an Explicit Congestion Notification (ECN) mechanism is set on the queue of each switch, and a network performance prediction model needs to predict completion times of different streams (start time and size) under different configurations.

As shown in fig. 7, flow F0 and flow F1 are configured to be transmitted from queue Q0 of server H0 to server H5 via path P0, flow F2 is transmitted from queue Q1 of server H0 to server H4 via path P1, flow F3 is configured to be transmitted from queue Q1 of server H1 to server H5 via path P2, where path P0 passes through queue Q0 of link L0 of switch S0, queue Q0 of link L0 of switch S4, and queue Q0 of link L0 of switch S2, and path P1 passes through queue Q1 of link L0 of switch S0, queue Q1 of link L0 of switch S4, and queue Q1 of link L1 of switch S2; path P2 passes through queue Q1 of link L0 of switch S0, queue Q1 of link L0 of switch S4, and queue Q1 of link L0 of switch S2.

Based on the above information, a heterogeneous relationship graph of the data center network can be constructed, as shown in fig. 8. To accurately model different levels of network feature information, the network is described using 5 different types of nodes, including switch nodes S0, S2, and S4, port (link) nodes L0 and L1, queue nodes Q0 and Q1, path nodes P1, P2 and P3, flow nodes F0, F1, F2, and F3. The connection between the nodes indicates that a direct influence relationship exists between the nodes. For example, the queue node has ECN threshold and priority weight, the port node has queue policy, and the queue node and the port node are connected together. The switch node has a control variable sharing the cache, and acts on the port node, and the switch node needs to be connected with the port node. The stream nodes are connected together with the path nodes they pass through. Thus, local relationships can be described by disjoint edges. In order to establish a global relationship, such as a route, a path node needs to be connected with a queue node through which it passes. Note that not all physical nodes in the data center network have to be instantiated, as some of the devices may not be streaming through, where they have no impact on the performance of interest.

First state information of the heterogeneous relationship graph shown in fig. 8 is obtained, and the first state information includes first state information of each switch node, first state information of each link (port) node, first state information of each queue node, first state information of each path node, and first state information of each flow node. In this example, the switch nodes, link nodes, and queue nodes all contain only static information, and the path and flow nodes contain both static and dynamic information. Specifically, the first state information of the switch node includes, but is not limited to, a cache size and a cache policy; the first state information of the link node includes but is not limited to bandwidth, propagation delay, queue scheduling policy, and cache management control factor; the first state information of the queue node includes but is not limited to queue scheduling weight, priority; the first state information of the path node comprises static information such as hop count and the like and dynamic information such as path flow size and the like; the first state information of the flow node includes static information including, but not limited to, a flow start time, a flow size, and dynamic information including, but not limited to, a remaining flow size, a flow lifetime. Since the dimensions of the first state information are different for different types of nodes, the third state information of each node is obtained by encoding the first state information of each node in the heterogeneous relationship diagram shown in fig. 8, and the dimensions of the third state information of all nodes in the heterogeneous relationship diagram shown in fig. 8 are consistent. In order to make the prediction of the performance index more accurate, the hidden information of the heterogeneous relationship diagram at the previous time and the third state information of the heterogeneous relationship diagram at the current time are processed, for example, spliced, to obtain the fourth state information of the heterogeneous relationship diagram at the current time, and a specific splicing process may refer to the related description of S402, which is not described herein.

In order to make the state information of each node in the heterogeneous relationship diagram shown in fig. 8 include the first state information of other nodes, multiple rounds of information transmission are performed on the fourth state information of the heterogeneous relationship diagram shown in fig. 8, so as to obtain the hidden state information of the heterogeneous relationship diagram shown in fig. 8. For each path node, aggregating the information of all the flows passing through the path by using an addition mode, then utilizing RNN to aggregate all the queue information passing through the path node, respectively converting the aggregated information into the hidden space of the state information of the path node through a conversion function corresponding to the path node, splicing the fourth state information of the path node, the converted flow information and the converted queue information together, and transmitting the spliced information to an updating function for updating the state information of the path node.

For each stream node, the fourth state information of the path where the stream is located is converted into the hidden space of the stream node state information through the conversion function corresponding to the stream node, then the converted state information of the path node and the fourth state information of the stream node are spliced together, and the information obtained by splicing is transmitted to the updating function corresponding to the stream node to update the state information of the path node.

For each queue node, aggregating the fourth state information of all paths passing through the queue through an aggregation function corresponding to the queue node, then converting the fourth state information of the link node and the aggregated state information of the path node which are related to the queue node into a hidden space where the queue node is located through a conversion function corresponding to the queue node, splicing the converted state information of the link node and the converted state information of the path node together, and transmitting the spliced information to an updating function corresponding to the queue node for updating the state information of the queue node.

For each link (port) node, aggregating the fourth state information of all queue nodes related to the link through an aggregation function corresponding to the link node, then converting the fourth state information of the switch node and the aggregated state information of the queue nodes into a hidden space where the link node is located through a conversion function corresponding to the link node, splicing the converted state information of the switch node and the converted state information of the queue nodes together, and transmitting the spliced information to an update function corresponding to the link node for updating the state information of the link node.

For each switch node, the fourth state information of the link node contained in the switch node is converted into the hidden space where the switch node is located through the conversion function corresponding to the switch node, then the converted state information of the link node and the fourth state information of the switch node are spliced together, and the information obtained through splicing is transmitted to the updating function corresponding to the switch node to update the state information of the switch node.

A round of information transfer can be accomplished as described above. After multiple rounds of the above information transmission processes, the hidden state information of each node in the heterogeneous relationship diagram shown in fig. 8 can be obtained, and the hidden state information of each node in the heterogeneous relationship diagram shown in fig. 8 includes the first state information of other nodes.

Decoding is performed by decoding the hidden state information of the heterogeneous relationship diagram shown in fig. 8 of the network to obtain the second state information of the heterogeneous relationship diagram shown in fig. 8. The second state information of the heterogeneous relationship diagram shown in fig. 8 includes performance indicators of the network shown in fig. 7, such as path delay, packet loss rate, and the like. The second status information of the heterogeneous relationship diagram shown in fig. 8 further includes a dynamic information change amount, which is a change amount of the dynamic information of the heterogeneous relationship diagram at the current time in the next time with respect to the dynamic information of the heterogeneous relationship diagram at the current time, such as a flow completion data amount, a flow survival time, and the like between two times. The first state information of the heterogeneous relationship diagram shown in fig. 8 includes static information and dynamic information of the heterogeneous relationship diagram at the current time, where all the switch node, the link node, the queue node, and the path node include only the static information, and the flow node includes both the static information and the dynamic information, specifically, the first state information of the switch node includes, but is not limited to, a cache size and a cache policy; the first state information of the link node includes but is not limited to bandwidth, propagation delay, queue scheduling policy, and cache management control factor; the first state information of the queue node includes but is not limited to queue scheduling weight, priority; the first state information of the path node includes, but is not limited to, hop count; the first state information of the stream node includes static information such as stream start time and stream size, and dynamic information such as residual stream size and stream lifetime. And obtaining the dynamic information of the heterogeneous relational graph at the current moment at the next moment according to the dynamic information and the dynamic information variation of the heterogeneous relational graph at the current moment, and obtaining the first state information of the heterogeneous relational graph at the current moment at the next moment based on the dynamic information and the static information of the heterogeneous relational graph at the current moment at the next moment. After the first state information of the heterogeneous relationship diagram at the current moment at the next moment is obtained, the performance index of the network shown in fig. 7 at the subsequent moment can be predicted. It is noted that in this example, this performance index is indirectly obtained for the FCT by the amount of dynamic state change, and specifically, the network performance prediction model predicts the FCT by the amount of data received between predicted times and updating the remaining flow size and lifetime at each time. Once the remaining flow size is below zero, the network performance prediction model considers the flow to be over and uses its lifetime as the predicted FCT.

In the above example, one possible usage scenario is that the traffic of the target network is not described by a flow, but is described as aggregated traffic information through a path, and the network performance prediction model needs to predict the change of the path performance over time under different configurations and path traffic. At this time, other nodes are not changed, and the path node simultaneously contains static information and dynamic information, wherein the static information comprises hop count and the like, and the dynamic information comprises path flow speed; the output path performance includes time delay, packet loss and the like. In this scenario, when the network performance prediction model is used for prediction and training, the heterogeneous relational graph only includes switch nodes, link (port) nodes, queue nodes and path nodes, and does not include flow nodes.

It should be understood that the scheme of the present application has no limitation to the network field of application, and may be used in fine-grained performance prediction of a data center network, and may also be applied in path-level performance prediction of a backbone network.

The heterogeneous relationship diagram describes the interaction relationship between different entities, and a similar method can also be applied to the simulation process in the physical field, wherein the entity constraint, number and relationship and the aggregation function may change, and in addition, a physical equation may need to be introduced to describe the change process of the entities over time, but the influence relationship between multiple entities can still be described by adopting a state transition process.

For different networks, the network structures, configuration information of devices in the networks, and the like may be different, and therefore, in order to obtain accurate performance indexes, the network performance prediction models used for different networks are different. Therefore, for different networks, corresponding network performance prediction models need to be trained.

How to train to obtain a network performance prediction model of the network is described below.

Referring to fig. 9, fig. 9 is a schematic flowchart of a network performance prediction model training method provided in the embodiment of the present application.

As shown in fig. 9, the method includes:

s901, the training device acquires a plurality of heterogeneous relationship graphs and fifth state information of each heterogeneous relationship graph.

The method comprises the steps that a plurality of heterogeneous relation graphs are obtained at different moments aiming at the same network, and each heterogeneous relation graph comprises a plurality of nodes and side information among the nodes; the state information of each heterogeneous relationship graph includes first state information of each node in the heterogeneous relationship graph.

Alternatively, the training device may be the training apparatus 120 in fig. 1a, or the prediction device 10 or the calculation device 20 in fig. 1 b.

It is noted that the nodes in the heterogeneous relationship graph include physical nodes and virtual nodes. Where a physical node represents a network entity with configuration information including, but not limited to, switches, servers, links (ports), queues. The virtual nodes identify entities related to network performance including, but not limited to, flows, paths.

The nodes contained in the heterogeneous relationship graph at different time instants may not be the same. Heterogeneous relationship graphs are used to characterize entities in a network that participate in the operation of the network at a time and the relationships between those entities. Since the entities participating in the network operation may not be the same at different times, the heterogeneous relationship diagrams at different times may not be the same. The side information between the entities is also the side information between the nodes.

It should be noted that, the relevant features of the heterogeneous relational diagram can be referred to the relevant description of the embodiment shown in fig. 2, and will not be described here.

S902, the training device conducts multiple rounds of training on the initial prediction model based on the multiple heterogeneous relationship graphs and the fifth state information of each heterogeneous relationship graph in the multiple heterogeneous relationship graphs to obtain a network performance prediction model.

In a possible embodiment, the predicting device performs multiple rounds of training on the initial prediction model based on the multiple heterogeneous relationship graphs and the fifth state information of each of the multiple heterogeneous relationship graphs to obtain the network performance prediction model, and specifically includes:

in one example, the prediction apparatus divides the plurality of heterogeneous relationship maps into a plurality of groups, each group including at least one heterogeneous relationship map; then, a plurality of sets of heterogeneous relationship maps and fifth state information thereof are used for multi-round training, and when the 1 st round of training is performed, the prediction device uses the fifth state information and hidden state information H 'of the heterogeneous relationship map of the 1 st set in the plurality of sets' ₁ Input to a Performance prediction model M ₁ Processing is carried out in the initial prediction model to obtain at least one sixth state information corresponding to the heterogeneous relational graph of the 1 st group, wherein the hidden state information is an empty set, and the sixth state information comprises a performance index and a dynamic information variation; inputting the performance index and the dynamic information variation included in at least one sixth state information corresponding to the heterogeneous relational graph of the group 1 into a loss function for calculation to obtain a loss value L ₁ (ii) a Based on the loss value L ₁ Adjusting performance prediction model M ₁ To obtain a performance prediction model M ₂ (ii) a … …, respectively; during the p-th round of training, the fifth state information and the hidden state information H 'of the heterogeneous relational graph of the p-th group in the multiple groups' _p Input to a Performance prediction model M _p The processing is carried out to obtain at least one sixth state information corresponding to the p-th group of heterogeneous relational graph, the sixth state information comprises a performance index and a dynamic information variation, and p is an integer larger than 0; at least one sixth shapeThe performance index and the dynamic information variation included in the state information are input into a loss function for calculation to obtain a loss value L _p (ii) a Based on the loss value L _p Adjusting performance prediction model M _p To obtain a performance prediction model M _p+1 (ii) a When p is greater than 1, hiding the state information H' _p The performance prediction model M is obtained according to the last heterogeneous relation graph in the p-1 group, and the specific mode refers to the following related description.

Based on the loss value L _p Adjusting performance prediction model M _p May be based on the loss value L _p Adjusting performance prediction model M in gradient descent manner _p The parameter (c) of (c).

In another example, after obtaining the preset number of heterogeneous relationship graphs and the fifth state information thereof, the prediction apparatus performs a round of training based on obtaining the preset number of heterogeneous relationship graphs and the fifth state information thereof, and the specific training process may be referred to the above related description.

Further, the initial prediction model includes an initial coding network, an initial information transfer network and an initial decoding network, and correspondingly, the performance prediction model M _p Comprising a coding network M _1p Information transfer network M _2p And decoding network M _3p . The fifth state information of each heterogeneous relationship graph includes fifth state information of each node in the heterogeneous relationship graph.

Performing the following processing on the fifth state information of each heterogeneous relationship diagram in the p-th group to obtain sixth state information corresponding to each heterogeneous relationship diagram:

by coding the network M _1p Processing the fifth state information of each node in the 1 st heterogeneous relationship graph in the p group to obtain seventh state information of the 1 st heterogeneous relationship graph; passing through an information transfer network M based on side information between a plurality of nodes included in a 1 st heterogeneous relationship graph _2p For the seventh state information of the 1 st heterogeneous relationship diagram and the hidden state information H of the 0 th heterogeneous relationship diagram ₁ Performing multiple rounds of information transmission operation to obtain the 1 st isomeric relationHidden state information H of system picture ₂ (ii) a By decoding the network M _3p Hidden state information H for 1 st heterogeneous relational graph ₂ Decoding to obtain sixth state information corresponding to the 1 st heterogeneous relation graph; wherein, the hidden state information H of the 0 th heterogeneous relational graph ₁ Is hidden state information H' _p ；

By encoding the network M based on the side information between the nodes comprised by the 2 nd heterogeneous relational graph _1p Processing the fifth state information of each node in the 2 nd heterogeneous relationship diagram in the p group to obtain seventh state information of the 2 nd heterogeneous relationship diagram; through a messaging network M _2p For the seventh state information of the 2 nd heterogeneous relational graph and the hidden state information H of the 1 st heterogeneous relational graph ₂ Performing multiple rounds of information transmission operation to obtain hidden state information H of the 2 nd heterogeneous relational graph ₃ (ii) a By decoding the network M _3p Hidden state information H for 2 nd heterogeneous relational graph ₃ Decoding to obtain sixth state information corresponding to the 2 nd heterogeneous relation graph;

……

encoding the network M based on the side information among the plurality of nodes included in the jth heterogeneous relational graph _1p Processing the fifth state information of each node in the jth heterogeneous relationship graph to obtain seventh state information of the jth heterogeneous relationship graph; through a messaging network M _2p For the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation to obtain hidden state information H of the jth heterogeneous relational graph _j+1 (ii) a By decoding the network N _3j Hidden state information H for jth heterogeneous relational graph _j+1 And decoding to obtain sixth state information corresponding to the jth heterogeneous relation graph.

In the above manner, the sixth state information of each heterogeneous relationship diagram in the pth group can be used. When p is 1, the network M is coded _1p For the initial coding of the network, the information transfer network M _2p For the initial information transfer network, decoding the network M _3p The network is initially decoded.

In one possible embodiment, the information delivery network M is passed based on side information between a plurality of nodes included in the jth heterogeneous relationship graph _2p For the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation to obtain hidden state information H of the jth heterogeneous relational graph _j+1 The method specifically comprises the following steps:

seventh state information based on jth heterogeneous relationship graph and hidden state information H of jth heterogeneous relationship graph-1 _j Obtaining eighth state information of a jth heterogeneous relation graph; the eighth state information of the jth heterogeneous relationship graph comprises eighth state information of each node in the jth heterogeneous relationship graph;

when the 1 st round of information transmission is carried out, determining the neighbor nodes of each node in the jth heterogeneous relationship graph based on the side information among the nodes of the jth heterogeneous relationship graph; according to the state information S of each node ₁ And state information S of its neighbor nodes ₁ Obtaining state information S of each node ₂ (ii) a Status information S of each node ₂ State information S of neighbor nodes including the node ₁ (ii) a Status information S ₁ Is the eighth status information;

when the 2 nd round of information transmission is carried out, determining the neighbor nodes of each node in the jth heterogeneous relationship graph based on the side information among the nodes of the jth heterogeneous relationship graph; according to the state information S of each node ₂ And state information S of its neighbor node ₂ Obtaining state information S of each node ₃ (ii) a Status information S of each node ₃ State information S of a neighbor node including the node ₂ ；

……

When the kth round of information transmission is carried out, state information S of neighbor nodes of each node in the jth heterogeneous relationship graph is obtained based on the jth heterogeneous relationship graph _k (ii) a State information S of neighbor nodes based on each node in jth heterogeneous relation graph _k And state information S of each node _k Obtaining state information S of each node _k+1 。

The eighth state information of the jth heterogeneous relationship graph comprises eighth state information of each node in the jth heterogeneous relationship graph, and when the kth round of information transfer is the last round of information transfer in multiple rounds of information transfer, the state information S of each node _i+1 Hidden state information H of jth heterogeneous relation graph as hidden state information of the node _j+1 Including hidden state information for each node in the jth heterogeneous relationship graph.

It should be understood that, since the neighbor nodes of each node in the 1 st heterogeneous relationship diagram have been determined based on the side information between the plurality of nodes of the jth heterogeneous relationship diagram at the time of the 1 st round of information transfer, the operation of "determining the neighbor nodes of each node in the jth heterogeneous relationship diagram based on the side information between the plurality of nodes of the jth heterogeneous relationship diagram" may not be performed at the time of the subsequent information transfer.

The aggregation operation may be implemented by an aggregation function, the conversion operation may be implemented by a conversion function, and the update operation may be implemented by an update function. The aggregation function, the conversion function, and the update function may each be implemented by a neural network, such as a graph neural network. For the heterogeneous relationship graph, nodes with the same type are included. For nodes of the same type, the corresponding aggregation functions may be the same, the corresponding conversion functions may be the same, and the corresponding update functions may be the same.

In one possible embodiment of the method according to the invention,hidden state information H of j-1 th heterogeneous relation graph _j Hidden state information of each node comprising the j-1 th heterogeneous relationship graph; seventh state information based on jth heterogeneous relationship graph and hidden state information H of jth heterogeneous relationship graph-1 _j Obtaining the eighth state information of the jth heterogeneous relationship diagram may specifically include:

and for any node Y in the heterogeneous relationship graph of the jth heterogeneous relationship graph, if the jth-1 heterogeneous relationship graph has the node Y, obtaining eighth state information of the node Y according to the seventh state information of the node Y and the hidden state information of the node Y. For example, the hidden state information of the node Y is represented by a 2-dimensional vector, the seventh state information of the node Y is represented by a 3-dimensional vector, and the eighth state information of the node Y is represented by a 4-dimensional vector obtained by the 2-dimensional vector representing the hidden state information of the node Y and the 3-dimensional vector representing the seventh state information of the node Y. And if the j-1 th heterogeneous relation graph does not have the node Y, determining the seventh state information of the node Y as the eighth state information of the node Y.

According to the method, the eighth state information of each node in the jth heterogeneous relationship graph can be obtained. The eighth state information of the jth heterogeneous relationship graph includes eighth state information of each node in the jth heterogeneous relationship graph.

According to the training method, a network performance prediction model of the network can be obtained.

Referring to fig. 10, fig. 10 is a schematic structural diagram of a prediction apparatus according to an embodiment of the present disclosure. As shown in fig. 10, the prediction apparatus 1000 includes:

an obtaining unit 1001, configured to obtain, for a target network, a current-time heterogeneous relationship graph and first state information of the current-time heterogeneous relationship graph, where the current-time heterogeneous relationship graph is used to represent multiple nodes participating in network operation in the current-time target network and relationships among the multiple nodes;

the prediction unit 1002 is configured to process the heterogeneous relationship diagram at the current time and the first state information of the heterogeneous relationship diagram at the current time through a network performance prediction model of the target network, to obtain second state information of the heterogeneous relationship diagram at the current time, where the second state information includes a performance index of the target network at the current time;

a determining unit 1003, configured to determine, based on the second state information and the first state information of the heterogeneous relationship diagram at the current time, first state information of the heterogeneous relationship diagram at the next time.

In a possible embodiment, the network performance prediction model of the target network includes an encoding network, an information delivery network, and a decoding network, and the prediction unit 1002 is specifically configured to:

In a possible embodiment, the heterogeneous relationship graph at the current time includes a plurality of nodes of the target network and side information between the plurality of nodes, where the side information between the plurality of nodes is used to represent a relationship between the plurality of nodes, and in an aspect that the heterogeneous relationship graph at the current time, the third state information of the heterogeneous relationship graph at the current time, and the hidden state information of the heterogeneous relationship graph at the previous time are subjected to multiple rounds of information transfer operations through the information transfer network to obtain the hidden state information of the heterogeneous relationship graph at the current time, the prediction unit 1002 is specifically configured to:

obtaining fourth state information of the heterogeneous relational graph at the current moment based on the third state information of the heterogeneous relational graph at the current moment and the hidden state information of the heterogeneous relational graph at the previous moment; the fourth state information of the heterogeneous relationship diagram at the current moment comprises the fourth state information of each node in the heterogeneous relationship diagram at the current moment; when the ith round of information transfer is carried out, the information transfer method is based on a plurality of nodesThe side information between the nodes obtains the state information S of the neighbor node of each node in the heterogeneous relation graph at the current moment _i (ii) a i is an integer greater than 0; state information S of neighbor nodes based on each node _i And state information S of each node _i Obtaining state information S of each node _i+1 (ii) a When i is equal to 1, the state information S of each node in the heterogeneous relationship graph at the current moment _i Fourth state information for each node; when the ith round of information transfer is the last round of information transfer in the multi-round of information transfer, the state information S of each node _i+1 And for the hidden state information of each node, the hidden state information of the heterogeneous relationship graph at the current moment comprises the hidden state information of each node in the heterogeneous relationship graph at the current moment.

In one possible embodiment, the state information S is based on the state information S of each node' S neighbor nodes _i And state information S of each node _i Obtaining state information S of each node _i+1 In an aspect, the prediction unit 1002 is specifically configured to:

In a possible embodiment, the first state information of the heterogeneous relationship diagram at the current time includes static information and dynamic information at the current time, the second state information further includes a state information variation, and the dynamic information variation is a variation between the dynamic information at the next time and the dynamic information at the current time, and the determining unit 1003 is specifically configured to:

It should be noted that the above units (the acquiring unit 1001, the predicting unit 1002, and the determining unit 1003 are used for executing the relevant steps of the above method, for example, the acquiring unit 1001 is used for executing the relevant content of S201, the predicting unit 1002 is used for executing the relevant content of S202, and the determining unit 1003 is used for executing the relevant content of S203.

In the present embodiment, the prediction apparatus 1000 is presented in the form of a unit. An "element" may refer to an application-specific integrated circuit (ASIC), a processor and memory that execute one or more software or firmware programs, an integrated logic circuit, and/or other devices that may provide the described functionality. Further, the above acquisition unit 1001, prediction unit 1002, and determination unit 1003 may be realized by the processor 1201 of the electronic apparatus illustrated in fig. 12.

Referring to fig. 11, fig. 11 is a schematic structural diagram of an exercise device according to an embodiment of the present disclosure. As shown in fig. 11, the training apparatus 1100 includes:

an obtaining unit 1101, configured to obtain a plurality of heterogeneous relationship graphs and fifth state information of each of the plurality of heterogeneous relationship graphs, where the plurality of heterogeneous relationship graphs are obtained at different times for a same network, and each of the plurality of heterogeneous relationship graphs is used to represent a plurality of nodes participating in network operation in the time when each of the plurality of heterogeneous relationship graphs is obtained and a relationship between the plurality of nodes in the network;

the training unit 1102 is configured to perform multiple rounds of training on the initial prediction model based on the multiple heterogeneous relationship graphs and the fifth state information of each of the multiple heterogeneous relationship graphs, so as to obtain a network performance prediction model.

In one possible embodiment, the training unit 1102 is specifically configured to:

dividing a plurality of heterogeneous relationship graphs into a plurality of groups, wherein each group comprises at least one heterogeneous relationship graph; in the process ofDuring the p-th round of training, the fifth state information and the hidden state information H 'of the heterogeneous relational graph of the p-th group in the multiple groups' _p Input into a Performance prediction model M _p The processing is carried out to obtain at least one sixth state information corresponding to the p-th group of heterogeneous relational graph, the sixth state information comprises a performance index and a dynamic information variation, and p is an integer larger than 0; inputting the performance index and the dynamic information variation included in the at least one sixth state information into a loss function for calculation to obtain a loss value L _p (ii) a Based on the loss value L _p Adjusting performance prediction model M _p To obtain a performance prediction model M _p+1 (ii) a When p is 1, the performance prediction model M _p Hidden state information H 'for the initial prediction model' _p Is an empty set; when p is greater than 1, hiding the state information H' _p Is obtained according to the last heterogeneous relation graph in the p-1 group, and when the p-th training is the last training, the performance prediction model M _p+1 Is a network performance prediction model.

In one possible embodiment, the initial prediction model comprises an initial coding network, an initial information transfer network and an initial decoding network, and the performance prediction model M _p Comprising a coding network M _1p Information transfer network M _2p And decoding network M _3p (ii) a Each heterogeneous relation graph comprises a plurality of nodes and side information among the nodes; the fifth state information of each heterogeneous relationship graph comprises fifth state information of each node in the heterogeneous relationship graph, and the fifth state information of the heterogeneous relationship graph of the p-th group in the multiple groups and hidden state information H' _p Input to a Performance prediction model M _p In order to obtain at least one sixth state information corresponding to the p-th group of heterogeneous relationship graphs, the training unit 1102 is specifically configured to:

by coding the network M _1p Processing the fifth state information of each node in the jth heterogeneous relationship graph to obtain a seventh state of the jth heterogeneous relationship graphState information; passing through an information transfer network M based on side information between a plurality of nodes included in a jth heterogeneous relationship graph _2p For the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation to obtain hidden state information H of the jth heterogeneous relational graph _j+1 (ii) a By decoding the network M _3p Hidden state information H for jth heterogeneous relational graph _j+1 Decoding to obtain sixth state information corresponding to the jth heterogeneous relation graph; when j is 1, hidden state information H of j-1 th heterogeneous relation graph _j Is hidden state information H' _p (ii) a When p is 1, the network M is coded _1p For the initial coding of the network, the information transfer network M _2p For the initial information transfer network, decoding the network M _3p The network is initially decoded.

In one possible embodiment, the information delivery network M is based on side information between a plurality of nodes comprised by the jth heterogeneous relational graph _2p For the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation to obtain hidden state information H of the jth heterogeneous relational graph _j+1 In this regard, the training unit 1102 is specifically configured to:

seventh state information based on jth heterogeneous relationship graph and hidden state information H of jth heterogeneous relationship graph-1 _j Obtaining eighth state information of a jth heterogeneous relation graph; the eighth state information of the jth heterogeneous relationship graph comprises eighth state information of each node in the jth heterogeneous relationship graph; when the kth round of information transmission is carried out, the state information S of the neighbor node of each node in the jth heterogeneous relation graph is obtained based on the side information among a plurality of nodes included in the jth heterogeneous relation graph _k (ii) a The neighbor node of each node is a node with an edge between the neighbor node and the node; k is an integer greater than 0; state information S of neighbor nodes based on each node in jth heterogeneous relation graph _k And state information S of each node _k Obtaining state information S of each node _k+1 (ii) a When k is 1, the state information S of the neighbor node _k Is a neighbor nodeEighth State information of the Point, State information of Each node S _k Eighth status information for each node; when the kth round of information transfer is the last round of information transfer in the multiple rounds of information transfer, the state information S of each node _i+1 Hidden state information of jth heterogeneous relational graph H as hidden state information of each node _j+1 Including hidden state information for each node in the jth heterogeneous relationship graph.

In one possible embodiment, the state information S of the neighbor nodes of each node in the jth-based heterogeneous relationship graph _k And state information S of each node _k Obtaining state information S of each node _k+1 In this regard, the training unit 1102 is specifically configured to:

The units (the obtaining unit 1101 and the training unit 1102) are configured to perform the relevant steps of the method. For example, the acquiring unit 1101 is configured to execute the relevant content of S901, and the training unit 1102 is configured to execute the relevant content of S902.

In this embodiment, the training apparatus 1100 is presented in the form of a unit. An "element" may refer to an application-specific integrated circuit (ASIC), a processor and memory that execute one or more software or firmware programs, an integrated logic circuit, and/or other devices that may provide the described functionality. Further, the above acquisition unit 1101 and training unit 1102 may be implemented by the processor 1301 of the electronic device shown in fig. 13.

The electronic device 1200 shown in fig. 12 may be implemented in the structure shown in fig. 12, and the electronic device 1200 includes at least one processor 1201, at least one memory 1202, and at least one communication interface 1203. The processor 1201, the memory 1202 and the communication interface 1203 are connected through the communication bus to complete communication therebetween.

The processor 1201 may be a general purpose Central Processing Unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to control the execution of programs according to the above schemes.

Communication interface 1203 is used for communicating with other devices or communication Networks, such as ethernet, Radio Access Network (RAN), Wireless Local Area Networks (WLAN), etc.

The Memory 1202 may be, but is not limited to, a Read-Only Memory (ROM) or other type of static storage device that can store static information and instructions, a Random Access Memory (RAM) or other type of dynamic storage device that can store information and instructions, an Electrically Erasable Programmable Read-Only Memory (EEPROM), a Compact Disc Read-Only Memory (CD-ROM) or other optical Disc storage, optical Disc storage (including Compact Disc, laser Disc, optical Disc, digital versatile Disc, blu-ray Disc, etc.), a magnetic Disc storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. The memory may be self-contained and coupled to the processor via a bus. The memory may also be integral to the processor.

The memory 1202 is used for storing application program codes for executing the above schemes, and is controlled by the processor 1201 to execute. The processor 1201 is configured to execute application program code stored in the memory 1202.

The memory 1202 stores code that may perform any of the model search methods provided above, such as:

the method comprises the steps that a heterogeneous relation graph at the current moment and first state information of the heterogeneous relation graph at the current moment are obtained for a target network, and the heterogeneous relation graph at the current moment is used for representing a plurality of nodes participating in network operation in the target network at the current moment and relations among the plurality of nodes; processing the heterogeneous relationship diagram at the current moment and the first state information of the heterogeneous relationship diagram at the current moment through a network performance prediction model of the target network to obtain second state information of the heterogeneous relationship diagram at the current moment, wherein the second state information comprises a performance index of the target network at the current moment; and determining the first state information of the heterogeneous relationship diagram at the current moment at the next moment based on the second state information and the first state information of the heterogeneous relationship diagram at the current moment.

The electronic device 1300 shown in fig. 13 may be implemented in the structure of fig. 13, and the electronic device 1300 includes at least one processor 1301, at least one memory 1302 and at least one communication interface 1303. The processor 1301, the memory 1302, and the communication interface 1303 are connected via the communication bus to complete communication therebetween.

The processor 1301 may be a general purpose Central Processing Unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to control the execution of programs according to the above schemes.

Communication interface 1303 allows communications with other devices or communication Networks, such as ethernet, Radio Access Network (RAN), Wireless Local Area Networks (WLAN), etc.

The Memory 1302 may be a Read-Only Memory (ROM) or other type of static storage device that can store static information and instructions, a Random Access Memory (RAM) or other type of dynamic storage device that can store information and instructions, an Electrically Erasable Programmable Read-Only Memory (EEPROM), a Compact Disc Read-Only Memory (CD-ROM) or other optical Disc storage, optical Disc storage (including Compact Disc, laser Disc, optical Disc, digital versatile Disc, blu-ray Disc, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited to these. The memory may be self-contained and coupled to the processor via a bus. The memory may also be integral to the processor.

The memory 1302 is used for storing application program codes for executing the above schemes, and the processor 1301 controls the execution. The processor 1301 is configured to execute application program code stored in the memory 1302.

The memory 1302 stores code that may perform any of the model search methods provided above, such as:

The present application further provides a computer storage medium, where the computer storage medium may store a program, and when the program is executed, the program includes some or all of the steps of any one of the network performance prediction methods described in the above method embodiments.

Embodiments of the present application further provide a computer storage medium, where the computer storage medium may store a program, and the program includes, when executed, some or all of the steps of the training method of any one of the network performance prediction models described in the above method embodiments.

It should be noted that, for simplicity of description, the above-mentioned method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present application is not limited by the order of acts described, as some steps may occur in other orders or concurrently depending on the application. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required in this application.

In the foregoing embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus may be implemented in other manners. For example, the above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one type of division of logical functions, and there may be other divisions when actually implementing, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of some interfaces, devices or units, and may be an electric or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable memory. Based on such understanding, the technical solution of the present application may be substantially implemented or a part of or all or part of the technical solution contributing to the prior art may be embodied in the form of a software product stored in a memory, and including several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method described in the embodiments of the present application. And the aforementioned memory comprises: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.

Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable memory, which may include: flash Memory disks, Read-Only memories (ROMs), Random Access Memories (RAMs), magnetic or optical disks, and the like.

The foregoing detailed description of the embodiments of the present application has been presented to illustrate the principles and implementations of the present application, and the above description of the embodiments is only provided to help understand the method and the core concept of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims

1. A method for predicting network performance, comprising:

acquiring a current-time heterogeneous relationship graph and first state information of the current-time heterogeneous relationship graph aiming at a target network, wherein the current-time heterogeneous relationship graph is used for representing a plurality of nodes participating in network operation in the current-time target network and relations among the nodes;

processing the current heterogeneous relationship diagram and the first state information of the current heterogeneous relationship diagram through a network performance prediction model of the target network to obtain second state information of the current heterogeneous relationship diagram, wherein the second state information comprises a current performance index of the target network;

and determining the first state information of the heterogeneous relationship diagram at the current moment at the next moment based on the second state information and the first state information of the heterogeneous relationship diagram at the current moment.

2. The method according to claim 1, wherein the network performance prediction model includes an encoding network, an information delivery network, and a decoding network, and the processing the current-time heterogeneous relationship diagram and the first state information of the current-time heterogeneous relationship diagram by the network performance prediction model of the target network to obtain the second state information of the current-time heterogeneous relationship diagram includes:

coding the first state information of the heterogeneous relationship diagram at the current moment through the coding network to obtain third state information of the heterogeneous relationship diagram at the current moment;

performing multiple rounds of information transmission operations on the heterogeneous relationship diagram at the current moment, the third state information of the heterogeneous relationship diagram at the current moment and the hidden state information of the heterogeneous relationship diagram at the previous moment through the information transmission network to obtain the hidden state information of the heterogeneous relationship diagram at the current moment;

and decoding the hidden state information of the heterogeneous relationship diagram at the current moment through the decoding network to obtain second state information of the heterogeneous relationship diagram at the current moment.

3. The method according to claim 2, wherein the heterogeneous relationship diagram at the current time includes a plurality of nodes of the target network and side information between the plurality of nodes, the side information between the plurality of nodes is used to represent a relationship between the plurality of nodes, and performing multiple rounds of information transfer operations on the heterogeneous relationship diagram at the current time, the third state information of the heterogeneous relationship diagram at the current time, and the hidden state information of the heterogeneous relationship diagram at a previous time through the information transfer network to obtain the hidden state information of the heterogeneous relationship diagram at the current time comprises:

obtaining fourth state information of the heterogeneous relational graph at the current moment based on the third state information of the heterogeneous relational graph at the current moment and the hidden state information of the heterogeneous relational graph at the previous moment; the fourth state information of the heterogeneous relationship diagram at the current moment comprises fourth state information of each node in the heterogeneous relationship diagram at the current moment;

when the ith round of information transmission is carried out, state information S of neighbor nodes of each node in the heterogeneous relational graph at the current moment is obtained based on the side information among the multiple nodes _i (ii) a I is an integer greater than 0;

state information S of neighbor nodes based on each node _i And state information S of each node _i Obtaining the state information S of each node _i+1 ；

When the i is equal to 1, the state information S of each node in the heterogeneous relationship graph at the current moment _i Fourth state information for said each node; when the ith round of information transfer is the last round of information transfer in the multiple rounds of information transfer, the state information S of each node _i+1 And for the hidden state information of each node, the hidden state information of the heterogeneous relationship graph at the current moment comprises the hidden state information of each node in the heterogeneous relationship graph at the current moment.

4. Method according to claim 3, characterized in that said state information S based on said neighbor nodes of each node _i And state information S of each node _i Obtaining the state information S of each node _i+1 The method comprises the following steps:

state information S of neighbor nodes to each node _i Carrying out a polymerization operation to obtainAggregated state information for each of the nodes;

converting the aggregated state information of each node to obtain converted state information of each node;

state information S of each node based on the converted state information of each node _i Updating to obtain the state information S of each node _i+1 。

5. The method according to any one of claims 1 to 4, wherein the first state information of the heterogeneous relationship diagram at the current time includes static information and dynamic information at the current time, the second state information further includes a state information change amount, the dynamic information change amount is a change amount between the dynamic information at the next time with respect to the dynamic information at the current time, and the determining the first state information of the heterogeneous relationship diagram at the next time based on the second state information and the first state information of the heterogeneous relationship diagram at the current time includes:

determining dynamic information variation from second state information of the heterogeneous relationship diagram at the current moment;

and obtaining first state information of the heterogeneous relationship graph at the current moment at the next moment according to the dynamic information variation, the dynamic information at the current moment and the static information.

6. The method of any of claims 1-5, wherein the plurality of nodes of the target network comprise: network devices, links, queues, paths, and flows.

7. A training method of a network performance prediction model is characterized by comprising the following steps:

acquiring a plurality of heterogeneous relationship graphs and fifth state information of each heterogeneous relationship graph in the plurality of heterogeneous relationship graphs, wherein the plurality of heterogeneous relationship graphs are acquired at different moments aiming at the same network, and each heterogeneous relationship graph is used for representing a plurality of nodes participating in network operation in the network at the moment of acquiring each heterogeneous relationship graph and relationships among the plurality of nodes;

and performing multiple rounds of training on the initial prediction model based on the multiple heterogeneous relationship graphs and the fifth state information of each heterogeneous relationship graph in the multiple heterogeneous relationship graphs to obtain a network performance prediction model.

8. The method of claim 7, wherein the performing multiple rounds of training on the initial prediction model based on the plurality of heterogeneous relationship graphs and the fifth state information of each of the plurality of heterogeneous relationship graphs to obtain the network performance prediction model comprises:

dividing the plurality of heterogeneous relationship graphs into a plurality of groups, each group comprising at least one heterogeneous relationship graph;

during the p-th round of training, the fifth state information and the hidden state information H 'of the heterogeneous relational graph of the p-th group in the multiple groups' _p Input to a Performance prediction model M _p The processing is performed to obtain at least one sixth state information corresponding to the p-th group of heterogeneous relational graphs, the sixth state information includes a performance index and a dynamic information variation, and p is an integer greater than 0;

inputting the performance index and the dynamic information variation included in the at least one sixth state information into a loss function for calculation to obtain a loss value L _p ；

Based on the loss value L _p Adjusting the performance prediction model M _p To obtain a performance prediction model M _p+1 (ii) a When p is 1, the performance prediction model M _p Is an initial prediction model, the hidden state information H' _p Is an empty set; when p is greater than 1, the hidden state information H' _p Is obtained according to the last heterogeneous relation graph in the p-1 group, and when the p-th training is the last training, the performance prediction model M _p+1 And predicting the network performance model.

9. The method of claim 8, whichCharacterized in that the initial prediction model comprises an initial coding network, an initial information transfer network and an initial decoding network, and the performance prediction model M _p Comprising a coding network M _1p Information transfer network M _2p And decoding network M _3p (ii) a Each heterogeneous relationship graph comprises a plurality of nodes and side information among the nodes; the fifth state information of each heterogeneous relationship graph comprises fifth state information of each node in the heterogeneous relationship graph, and the fifth state information and hidden state information H 'of the p-th group of heterogeneous relationship graphs in the multiple groups' _p Input to a Performance prediction model M _p The processing to obtain at least one sixth state information corresponding to the p-th group of heterogeneous relationship maps includes:

by coding the network M _1p Processing the fifth state information of each node in the jth heterogeneous relationship graph to obtain seventh state information of the jth heterogeneous relationship graph;

based on the side information among a plurality of nodes included in the jth heterogeneous relational graph, passing through an information transfer network M _2p For the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation to obtain hidden state information H of the jth heterogeneous relationship diagram _j+1 ；

By decoding the network M _3p Hidden state information H for the jth heterogeneous relational graph _j+1 Decoding to obtain sixth state information corresponding to the jth heterogeneous relation graph;

when j is 1, hidden state information H of j-1 th heterogeneous relationship graph _j Is the hidden state information H' _p (ii) a When p is 1, the coding network M _1p For the initial coding network, the information transfer network M _2p For the initial information transfer network, the decoding network M _3p Is the initiationAnd decoding the network.

10. The method according to claim 9, wherein the information passing network M is based on side information between a plurality of nodes included in the jth heterogeneous relationship graph _2p For the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation to obtain hidden state information H of the jth heterogeneous relationship diagram _j+1 The method comprises the following steps:

based on the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Obtaining eighth state information of the jth heterogeneous relation graph; the eighth state information of the jth heterogeneous relationship graph comprises eighth state information of each node in the jth heterogeneous relationship graph;

when the kth round of information transmission is carried out, acquiring the state information S of the neighbor node of each node in the jth heterogeneous relationship graph based on the side information among the plurality of nodes included in the jth heterogeneous relationship graph _k (ii) a The neighbor node of each node is a node with an edge between the neighbor node and the node; k is an integer greater than 0;

state information S of neighbor nodes based on each node in the jth heterogeneous relationship graph _k And state information S of each node _k Obtaining the state information S of each node _k+1 ；

When the k is 1, the state information S of the neighbor node _k As the eighth state information of the neighbor node, the state information S of each node _k Eighth state information for said each node; when the k-th information transmission is the last information transmission in the multi-round information transmission, the state information S of each node _i+1 Hidden state information H of the jth heterogeneous relational graph is hidden state information of each node _j+1 And hidden state information of each node in the jth heterogeneous relationship graph is included.

11. According toThe method of claim 10, wherein the state information S based on neighbor nodes of each node in the jth heterogeneous relationship graph _k And state information S of each node _k Obtaining the state information S of each node _k+1 The method comprises the following steps:

state information S of neighbor nodes to each node _k Performing aggregation operation to obtain the aggregated state information of each node;

updating the state information of each node based on the converted state information of each node to obtain the state information S of each node _k+1 。

12. The method according to any of claims 7-11, wherein the plurality of nodes comprises: network devices, links, queues, paths, and flows.

13. A prediction apparatus, comprising:

the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a heterogeneous relationship graph at the current time and first state information of the heterogeneous relationship graph at the current time aiming at a target network, and the heterogeneous relationship graph at the current time is used for representing a plurality of nodes participating in network operation in the target network at the current time and relations among the nodes;

the prediction unit is configured to process the current heterogeneous relationship diagram and first state information of the current heterogeneous relationship diagram through a network performance prediction model of the target network to obtain second state information of the current heterogeneous relationship diagram, where the second state information includes a current performance index of the target network;

14. The apparatus of claim 13, wherein the network performance prediction model comprises an encoding network, an information delivery network, and a decoding network, and wherein the prediction unit is specifically configured to predict the network performance

15. The apparatus according to claim 14, wherein the heterogeneous relationship diagram at the current time includes a plurality of nodes of the target network and side information between the plurality of nodes, the side information between the plurality of nodes is used to represent a relationship between the plurality of nodes, and in the aspect of performing multiple rounds of information transfer operations on the heterogeneous relationship diagram at the current time, the third state information of the heterogeneous relationship diagram at the current time, and the hidden state information of the heterogeneous relationship diagram at a previous time through the information transfer network to obtain the hidden state information of the heterogeneous relationship diagram at the current time, the prediction unit is specifically configured to:

when the ith round of information transmission is carried out, state information S of neighbor nodes of each node in the heterogeneous relationship graph at the current moment is obtained based on the side information among the nodes _i (ii) a The neighbor node of each node is a node with an edge between the neighbor node and the node; i is an integer greater than 0;

16. The apparatus according to claim 15, wherein said state information S is based on said neighbor nodes of each node _i And state information S of each node _i Obtaining the state information S of each node _i+1 In an aspect, the prediction unit is specifically configured to:

state information S of neighbor nodes to each node _i Performing aggregation operation to obtain the aggregated state information of each node;

17. The apparatus according to claims 13-16, wherein the first state information of the heterogeneous relationship diagram at the current time includes static information and dynamic information at the current time, the second state information further includes a state information change amount, the dynamic information change amount is a change amount between the dynamic information at the next time and the dynamic information at the current time, the determining unit is specifically configured to:

18. The apparatus of any of claims 13-17, wherein the plurality of nodes of the target network comprise: network devices, links, queues, paths, and flows.

19. An exercise device, comprising:

an obtaining unit, configured to obtain a plurality of heterogeneous relationship graphs and fifth state information of each of the plurality of heterogeneous relationship graphs, where the plurality of heterogeneous relationship graphs are obtained at different times for a same network, and each of the plurality of heterogeneous relationship graphs is used to represent a plurality of nodes participating in network operation in the network at the time of obtaining each of the plurality of heterogeneous relationship graphs and a relationship between the plurality of nodes;

and the training unit is used for carrying out multiple rounds of training on the initial prediction model based on the multiple heterogeneous relationship graphs and the fifth characteristic information of each heterogeneous relationship graph in the multiple heterogeneous relationship graphs so as to obtain a network performance prediction model.

20. The apparatus according to claim 19, wherein the training unit is specifically configured to:

Based on the loss value L _p Adjusting the performance prediction model M _j To obtain a performance prediction model M _p+1 (ii) a When p is 1, the performance prediction model M _p Is an initial prediction model, the hidden state information H' _p Is an empty set; when p is greater than 1, the hidden state information H' _p Is obtained according to the last heterogeneous relation graph in the p-1 group, and when the p-th training is the last training, the performance prediction model M _p+1 And predicting the network performance model.

21. The apparatus of claim 20, wherein the initial prediction model comprises an initial coding network, an initial information transfer network, and an initial decoding network, and wherein the performance prediction model M is _p Comprising a coding network M _1p Information transfer network M _2p And decoding network M _3p (ii) a Each heterogeneous relationship graph comprises a plurality of nodes and side information among the nodes; the fifth state information of each heterogeneous relationship graph comprises fifth state information of each node in the heterogeneous relationship graph, and the fifth state information and hidden state information H 'of the p-th heterogeneous relationship graph in the multiple groups' _p Input to a Performance prediction model M _p To obtain the correspondence of the p-th group of heterogeneous relationship graphsIn an aspect of the at least one sixth status information, the training unit is specifically configured to:

processing the fifth state information of the jth heterogeneous relationship diagram in the pth group as follows to obtain sixth state information corresponding to the jth heterogeneous relationship diagram;

passing through an information transfer network M based on side information between a plurality of nodes included in the jth heterogeneous relationship graph _2p For the seventh state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation to obtain hidden state information H of the jth heterogeneous relationship diagram _j+1 ；

By decoding the network M _3p Hidden state information H for the jth heterogeneous relational graph _j+1 Decoding to obtain sixth state information corresponding to the jth heterogeneous relationship diagram

When j is 1, hidden state information H of j-1 th heterogeneous relationship graph _j Is the hidden state information H' _p (ii) a When p is 1, the coding network M _1p For the initial coding network, the information transfer network M _2p For the initial information transfer network, the decoding network M _3p A network is decoded for the initial.

22. The apparatus according to claim 21, wherein the information passing network M is configured to pass through based on side information between a plurality of nodes included in the jth heterogeneous relationship graph _2p The eighth state information of the jth heterogeneous relationship diagram and the hidden state information H of the jth-1 heterogeneous relationship diagram _j Performing multiple rounds of information transmission operation to obtain hidden state information H of the jth heterogeneous relationship diagram _j+1 In an aspect, the training unit is specifically configured to:

seventh heterogeneous relationship graph based on the jthHidden state information H of state information and j-1 heterogeneous relation graph _j Obtaining eighth state information of the jth heterogeneous relation graph; the eighth state information of the jth heterogeneous relationship graph comprises eighth state information of each node in the jth heterogeneous relationship graph;

acquiring state information S of neighbor nodes of each node in the jth heterogeneous relationship graph based on the jth heterogeneous relationship graph during the kth round of information transmission _k (ii) a The neighbor node of each node is a node with an edge between the neighbor node and the node;

state information S of neighbor nodes based on each node in the jth heterogeneous relationship graph _k And state information S of each node _k Obtaining the state information S of each node _k+1 (ii) a k is an integer greater than 0;

23. The apparatus according to claim 22, wherein the state information S of neighbor nodes based on each node in the jth heterogeneous relationship graph _k And state information S of each node _k Obtaining the state information S of each node _k+1 In an aspect, the training unit is specifically configured to:

24. The apparatus according to any of claims 19-23, wherein the plurality of nodes comprises: network devices, links, queues, paths, and flows.

25. An electronic device comprising a processor and a memory, wherein the processor is coupled to the memory, wherein the memory is configured to store program code and the processor is configured to invoke the program code to implement the method of any of claims 1 to 12.

26. A computer storage medium comprising computer instructions that, when executed on an electronic device, cause the electronic device to perform the method of any of claims 1-12.