CN116055378A - Training method and device for traffic scheduling strategy generation model - Google Patents
Training method and device for traffic scheduling strategy generation model Download PDFInfo
- Publication number
- CN116055378A CN116055378A CN202310032821.2A CN202310032821A CN116055378A CN 116055378 A CN116055378 A CN 116055378A CN 202310032821 A CN202310032821 A CN 202310032821A CN 116055378 A CN116055378 A CN 116055378A
- Authority
- CN
- China
- Prior art keywords
- information
- preset
- flow
- link delay
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000012549 training Methods 0.000 title claims abstract description 95
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000004891 communication Methods 0.000 claims abstract description 98
- 239000011159 matrix material Substances 0.000 claims abstract description 44
- 238000005259 measurement Methods 0.000 claims abstract description 19
- 238000010586 diagram Methods 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 6
- 238000005206 flow analysis Methods 0.000 claims description 4
- 230000003993 interaction Effects 0.000 claims description 4
- 238000003062 neural network model Methods 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000032683 aging Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000007723 transport mechanism Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/12—Shortest path evaluation
- H04L45/121—Shortest path evaluation by minimising delays
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/12—Shortest path evaluation
- H04L45/124—Shortest path evaluation using a combination of metrics
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L45/00—Routing or path finding of packets in data switching networks
- H04L45/38—Flow based routing
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/24—Traffic characterised by specific attributes, e.g. priority or QoS
Landscapes
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a training method and a training device for a flow scheduling strategy generation model, wherein the method comprises the following steps: acquiring flow information of each network node and link delay information of communication links between each network node; acquiring a target link delay prediction model according to the flow information and the link delay information; generating training data according to preset service quality information, a preset traffic matrix and a target link delay prediction model, wherein the preset service quality information is used for representing service quality information of preset service, and the training data represents traffic scheduling strategies aiming at the preset service in different network states; training the initial flow scheduling strategy generation model according to the training data to obtain a target flow scheduling strategy generation model, wherein the target flow scheduling measurement generation model is used for generating flow scheduling strategies for all network nodes.
Description
Technical Field
The present invention relates to the field of computer technologies, and in particular, to a training method for a traffic scheduling policy generation model, a traffic scheduling method, a traffic scheduling device, an electronic device, and a computer readable medium.
Background
A software defined network (SDN, software Defined Network) is an implementation manner of network virtualization, which separates a control plane and a data plane of network equipment, so that flexible control of network traffic is realized, the network becomes more intelligent, and a good platform is provided for innovation of a core network and application.
In the environment of rapid flow growth and more stringent service quality requirements, a new solution idea can be provided for network congestion based on SDN, but the integration of a software defined network and a flow scheduling technology is difficult. Currently, when traffic scheduling is performed, modeling a network is generally relied on, and when the network changes, the network needs to be modeled again, which may increase the resource consumption of a Data Center (Data Center), and it is often difficult to accurately acquire a traffic matrix in a practical environment.
It can be seen that the existing traffic scheduling method is difficult to adapt to the uncertainty of the network traffic, may have aging problems, and cannot solve the congestion situation which may exist in the global network.
Disclosure of Invention
Therefore, the invention provides a training method of a traffic scheduling strategy generation model, which aims to solve the problems that network traffic can not be scheduled in time and network global congestion can not be solved possibly existing in the related technology.
In order to achieve the above object, a first aspect of the present invention provides a training method of a traffic scheduling policy generation model, the method including:
acquiring flow information of each network node and link delay information of communication links between the network nodes;
acquiring a target link delay prediction model according to the flow information and the link delay information;
generating training data according to preset service quality information, a preset flow matrix and the target link delay prediction model, wherein the preset service quality information is used for representing service quality information of preset service, and the training data represents flow scheduling strategies aiming at the preset service in different network states;
training an initial flow scheduling strategy generation model according to the training data to obtain a target flow scheduling strategy generation model, wherein the target flow scheduling measurement generation model is used for generating flow scheduling strategies for all network nodes.
Optionally, the obtaining a target link delay prediction model according to the traffic information and the link delay information includes:
generating a plurality of communication links according to the traffic information and the link delay information, wherein the communication links are formed by a first network node and a second network node, and the first network node and the second network node are any different network nodes in the network nodes;
Constructing a communication link diagram according to the plurality of communication links, wherein the communication link diagram takes each network node in the plurality of communication links as a diagram node, and takes at least traffic information and link delay among different communication links as connecting edges among the corresponding diagram nodes;
training an initial link delay prediction model by using the communication link graph to obtain the target link delay prediction model, wherein the target link delay prediction model is used for predicting link delay between all network nodes.
Optionally, the generating training data according to the preset service quality information, the preset traffic matrix and the target link delay prediction model includes:
acquiring a target communication path set among the network nodes according to the preset service quality information, the preset traffic matrix and the target link delay prediction model, wherein the target communication path set consists of communication paths meeting preset conditions in communication paths for forwarding traffic of the preset service by the network nodes in a network state represented by the preset traffic matrix;
And generating the training data according to the preset service quality information, the preset flow matrix and the target communication path set.
Optionally, the obtaining a target communication path set between the network nodes according to the preset service quality information, the preset traffic matrix and the target link delay prediction model includes:
generating a plurality of flow matrix combinations according to the flow type of the preset service corresponding to the service quality information and the preset flow matrix;
and calculating communication paths which correspond to the plurality of traffic matrix combinations and meet the preset conditions by using a preset routing algorithm so as to obtain the target communication path set.
Optionally, training the initial traffic scheduling policy generation model according to the training data to obtain a target traffic scheduling policy generation model, including:
acquiring the initial flow scheduling strategy generation model constructed based on the BP neural network model structure;
inputting the training data into the initial flow scheduling measurement model to obtain a predicted flow scheduling strategy;
calculating an error value between a predicted link delay corresponding to the predicted traffic scheduling policy and an expected link delay corresponding to a communication path in the training data;
And adjusting the parameters of the initial flow scheduling strategy by taking the error value as a loss value to obtain an initial flow scheduling strategy generation model meeting a preset convergence condition as the target flow scheduling strategy generation model.
Optionally, the acquiring the traffic information of each network node includes:
analyzing the communication data among the network nodes by using a preset network flow analysis tool to obtain initial flow information of the network nodes;
classifying the initial flow information according to the service type corresponding to the initial flow information;
and filtering and normalizing the initial flow information after the classification processing to obtain the flow information.
In order to achieve the above object, a second aspect of the present invention provides a traffic scheduling method, including:
acquiring network state information of a current network, wherein the network state information at least comprises a service type of a service to be subjected to flow scheduling currently and a predicted link delay between communication links in the current network;
inputting the network state information into a target flow scheduling measurement generation model to obtain a preset flow scheduling strategy;
And carrying out flow scheduling based on the preset flow scheduling strategy.
In order to achieve the above object, a third aspect of the present invention further provides a training device for a traffic scheduling policy generation model, the device including:
the information acquisition module is used for acquiring the flow information of each network node and acquiring the link delay information of the communication links between the network nodes;
the link delay prediction model acquisition module is used for acquiring a target link delay prediction model according to the flow information and the link delay information;
the generation module is used for generating training data according to preset service quality information, a preset flow matrix and the target link delay prediction model, wherein the preset service quality information is used for representing service quality information of preset service, and the training data represents flow scheduling strategies aiming at the preset service in different network states;
the training module is used for training the initial flow scheduling strategy generation model according to the training data to obtain a target flow scheduling strategy generation model, wherein the target flow scheduling measurement generation model is used for generating flow scheduling strategies for all the network nodes.
In order to achieve the above object, a fourth aspect of the present invention further provides a traffic scheduling device, including:
the system comprises a state information acquisition module, a traffic scheduling module and a traffic scheduling module, wherein the state information acquisition module is used for acquiring network state information of a current network, wherein the network state information at least comprises a service type of a service to be subjected to traffic scheduling currently and a predicted link delay among communication links in the current network;
the flow scheduling strategy obtaining module is used for inputting the network state information into the target flow scheduling measurement generation model to obtain a preset flow scheduling strategy;
and the flow scheduling module is used for carrying out flow scheduling based on a preset flow scheduling strategy.
In order to achieve the above object, a fifth aspect of the present invention further provides an electronic device, including:
one or more processors;
a memory having one or more programs stored thereon which, when executed by the one or more processors, cause the one or more processors to implement the method of any of the first or second aspects of the present invention;
one or more I/O interfaces coupled between the processor and the memory configured to enable information interaction of the processor with the memory.
In order to achieve the above object, a sixth aspect of the present invention further provides a computer readable medium having stored thereon a computer program which, when executed by a processor, implements the method of any one of the first or second aspects of the present invention.
The invention has the following advantages: according to the embodiment of the invention, the flow information of each network node in the current network and the link delay information of the communication links between each network node are obtained; according to the flow information and the link delay information, a target link delay prediction model capable of accurately predicting the link delay between communication links is obtained; and generating training data according to the preset service quality information, the preset flow matrix and the target link delay prediction model, training the initial flow scheduling strategy generation model by using the training data to obtain a target flow scheduling strategy generation model, and carrying out flow scheduling on the current network state timely and globally based on the target flow scheduling measurement generation model.
Drawings
The accompanying drawings are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate the invention and together with the description serve to explain, without limitation, the invention.
Fig. 1 is a flow chart of a training method of a flow scheduling policy generation model according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of obtaining a target link delay prediction model according to an embodiment of the present invention;
fig. 3 is a flow chart of a flow scheduling method according to an embodiment of the present invention;
FIG. 4 is a block diagram of a training device for generating a model of a traffic scheduling policy according to an embodiment of the present invention;
fig. 5 is a block diagram of a flow scheduling device according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The following describes specific embodiments of the present invention in detail with reference to the drawings. It should be understood that the detailed description and specific examples, while indicating and illustrating the invention, are not intended to limit the invention.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above-described drawings are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order; in addition, the embodiments of the present invention and the features in the embodiments may be arbitrarily combined with each other without collision.
As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
The terms "comprises," "comprising," and/or "including" when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and the present invention and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
Aiming at the problems that the flow scheduling method in the related technology is difficult to adapt to the uncertainty of network flow, may have aging problems and cannot solve the congestion situation possibly existing in the whole network, the embodiment of the invention provides a training method of a flow scheduling strategy generation model, so as to train and obtain a target flow scheduling strategy generation model, and the flow scheduling is timely and globally performed on the current network state based on the target flow scheduling strategy generation model. Fig. 1 is a flow chart of a training method of a traffic scheduling policy generation model according to an embodiment of the present invention. The method may be implemented by an electronic device, which may be a server, for example, may be a physical server, for example, may be a blade server, a rack-mounted server, or may be a virtual server, for example, may be a server cluster deployed in the cloud; of course, the electronic device may be a terminal device, which is not particularly limited herein.
As shown in fig. 1, the training method of the traffic scheduling policy generation model provided in the embodiment of the present invention may include the following steps S101 to S104, which are described in detail below.
Step S101, acquiring traffic information of each network node, and acquiring link delay information of communication links between each network node.
The network node in the embodiment of the invention can be any one or more network nodes used for forwarding traffic data in a data center, and the network node can be, for example, a routing device.
The traffic information may be information indicating traffic conditions of each network node acquired based on network protocols such as NetStream protocol and sFlow.
In some embodiments, the acquiring the traffic information of each network node in step S101 may be: analyzing communication data among all network nodes by using a preset network flow analysis tool to obtain initial flow information of all network nodes; classifying the initial flow information according to the service type corresponding to the initial flow information; and filtering and normalizing the classified initial flow information to obtain flow information.
The preset network traffic analysis tool may be a tool based on network protocols such as NetStream protocol and sFlow, for example, may be network packet analysis software, windshare and tcpdu.
The initial traffic information may be traffic information captured by the preset network traffic analysis tool according to voice, text, video, etc.
After the initial traffic information is obtained, in order to facilitate the subsequent modeling of the network to perform traffic scheduling on the network with pertinence for the traffic type of the service application, the initial traffic information may be classified, for example, the obtained initial traffic information of the voice, text and video type may be classified by using one-hot coding, and then the classified initial traffic information may be filtered and normalized to obtain the traffic information.
In the embodiment of the invention, the initial flow information is classified according to the voice, text and video types, and in actual implementation, the flow information can be classified in other modes according to the needs; when the one-hot encoding is used to classify the initial traffic information, the voice type may be represented by [1, 0], the text type may be represented by [0,1,0], and the video type may be represented by [0,1], without limitation.
In addition, when filtering the initial flow information after the classification processing, the obvious numerical value abnormality or unnecessary data can be removed according to the type of the initial flow information, so that the data processing amount of the subsequent processing is reduced and the accuracy of a model obtained based on the training of the initial flow information is improved; the normalization process may be a unified quantization process for the initial flow information after the filtering process, for example, the unified normalization process for the value of the initial flow information is a value in the range of the [0,1] interval, so as to facilitate the subsequent feature analysis thereof.
In some embodiments, the obtaining the link delay information of the communication link between the network nodes in step S101 may be measuring the link delay between the network nodes at the host, for example, the link delay between the network nodes may be measured by using the P4 (Programming Protocol-independent Packet Processors) +in-band network telemetry (INT, INT In band Network Telemetry) method.
Step S102, a target link delay prediction model is obtained according to the flow information and the link delay information.
Please refer to fig. 2, which is a flowchart illustrating a process of obtaining a target link delay prediction model according to an embodiment of the present invention. As shown in fig. 2, in some embodiments, the obtaining the target link delay prediction model according to the traffic information and the link delay information includes:
step S201, generating a plurality of communication links according to the traffic information and the link delay information, wherein the communication links are formed by a first network node and a second network node, and the first network node and the second network node are any different network nodes in the network nodes.
Step S202, a communication link diagram is constructed according to a plurality of communication links, wherein each network node in the communication links is used as a diagram node, and at least traffic information and link delay between different communication links are used as connecting edges between the corresponding diagram nodes.
Step S203, training the initial link delay prediction model by using the communication link graph to obtain a target link delay prediction model, wherein the target link delay prediction model is used for predicting the link delay between each network node.
In an embodiment of the present invention, the initial link delay prediction model may be a graph neural network (GNN, graph Neural Network) model.
That is, in the embodiment of the present invention, the traffic information of each network node in the data center at different time instants may be obtained, and the link delay of the communication link formed by each network node at different time instants may be measured; and then, taking each network node as a graph node, taking traffic information and link delay between each graph node as a connecting edge to construct a communication link graph, and training an initial link delay prediction model by using the communication link graph so as to obtain a target link delay prediction model capable of accurately predicting the link delay between each communication link in the network, wherein the link delay of the network can be predicted based on the target link delay prediction model, so that the congestion condition of the whole network can be known globally, and a traffic scheduling strategy can be accurately generated based on the predicted link delay later so as to solve the congestion problem of the network.
In some embodiments, considering that traffic conditions in a data center network are complex and variable, if a target link delay prediction model is not continuously optimized after the target link delay prediction model is obtained by training, the target link delay prediction model may have a problem that the link delay between communication links in the network cannot be accurately predicted with low precision due to time variation, and therefore, in the embodiment of the present invention, a method for continuously optimizing the target link delay prediction model is further provided, which specifically includes: after training at a first moment to obtain a first target link delay prediction model, acquiring original prediction precision of the first target link delay prediction model, and backing up the first target link delay prediction model to obtain a second target link delay prediction model; continuously training the first target link delay prediction model according to link delay data in a preset time interval, and acquiring the current prediction precision of the continuously trained first target link delay prediction model; and under the condition that the current prediction precision is lower than the original prediction precision, replacing the continuously trained first target link delay prediction model by using a second target link delay prediction model to carry out subsequent link delay prediction processing.
That is, in the embodiment of the present invention, after the target link delay prediction model is obtained through the training in the steps S201 to S203, for example, after the model 1, the model 1 may be backed up to obtain the model 2, and the original measurement accuracy of the model 1 may be obtained; after that, the model 1 is put into use on line, and the model 1 is continuously trained by periodically using the latest link delay data in the network, if the current prediction accuracy of the continuously trained model 1 is lower than that of the original model, namely, the model 2 in a continuous period of time, in order to improve the accuracy of model prediction, the model 2 subjected to continuous optimization used on the current line can be replaced by the model 2 backed up, and the model 2 can be optimized by using the latest link delay data.
Of course, the above is only one embodiment provided in the embodiments of the present invention for continuously optimizing the target link delay prediction model, and other manners may be used to continuously optimize the target link delay prediction model in practical implementation, which is not limited herein.
Step S103, generating training data according to preset service quality information, a preset flow matrix and a target link delay prediction model, wherein the preset service quality information is used for representing service quality information of preset service, and the training data represents flow scheduling strategies aiming at the preset service in different network states.
In some embodiments, generating training data according to the preset quality of service information, the preset traffic matrix, and the target link delay prediction model includes: acquiring a target communication path set among all network nodes according to preset service quality information, a preset traffic matrix and a target link delay prediction model, wherein the target communication path set consists of communication paths meeting preset conditions in communication paths for forwarding traffic of preset services under the network state represented by the preset traffic matrix of all network nodes; and generating training data according to the preset service quality information, the preset traffic matrix and the target communication path set.
In this embodiment, the obtaining, according to the preset service quality information, the preset traffic matrix and the target link delay prediction model, the target communication path set between the network nodes includes: generating a plurality of flow matrix combinations with different combinations according to the flow type of the preset service and the preset flow matrix corresponding to the service quality information; and calculating communication paths which correspond to the traffic matrix combinations of the different combinations and meet preset conditions by using a preset routing algorithm so as to obtain a target communication path set.
The preset traffic matrix may be a traffic matrix corresponding to different services, which is obtained by measuring or estimating traffic in the network in advance.
That is, in the embodiment of the present invention, an optimal communication path set between network nodes may be calculated according to requirements of quality of service information (Qos, quality of Service) of different services and different traffic matrices, for example, a heuristic routing algorithm may be used to calculate an optimal communication path set between network nodes; and then, constructing training data according to all combinations of Qos requirements of different services and different traffic matrixes and optimal path samples thereof.
Step S104, training the initial flow scheduling strategy generation model according to training data to obtain a target flow scheduling strategy generation model, wherein the target flow scheduling measurement generation model is used for generating flow scheduling strategies for all network nodes.
After the training data is constructed, the training data may be used to train the initial traffic scheduling policy generation model to obtain the target traffic scheduling policy generation model.
In some embodiments, training the initial traffic scheduling policy generation model according to the training data to obtain the target traffic scheduling policy generation model includes: acquiring an initial flow scheduling strategy generation model constructed based on a BP neural network model structure; inputting training data into an initial flow scheduling measurement model to obtain a predicted flow scheduling strategy; calculating an error value between a predicted link delay corresponding to the predicted traffic scheduling policy and an expected link delay corresponding to the communication path in the training data; and adjusting parameters of the initial flow scheduling strategy by taking the error value as a loss value to obtain an initial flow scheduling strategy generation model meeting preset convergence conditions as a target flow scheduling strategy generation model.
Therefore, according to the training method of the traffic scheduling strategy generation model provided by the embodiment of the invention, the traffic information of each network node in the current network and the link delay information of the communication links between each network node are obtained; according to the flow information and the link delay information, a target link delay prediction model capable of accurately predicting the link delay between communication links is obtained; and generating training data according to the preset service quality information, the preset flow matrix and the target link delay prediction model, training the initial flow scheduling strategy generation model by using the training data to obtain a target flow scheduling strategy generation model, and carrying out flow scheduling on the current network state timely and globally based on the target flow scheduling measurement generation model.
Corresponding to the above embodiment, the embodiment of the present invention further provides a flow scheduling method, please refer to fig. 3, which is a flow chart of the flow scheduling method provided by the embodiment of the present invention. The method can be applied to the software defined network to timely and comprehensively schedule the flow according to the current state of the network in the Data Center (Data Center) by the software defined network, so that the problems of high network communication delay and network congestion possibly existing are avoided.
As shown in fig. 3, the flow scheduling method provided by the embodiment of the present invention includes the following steps S301 to S303:
step S301, obtaining network status information of the current network, where the network status information includes at least a service type of a service to be currently traffic scheduled and a predicted link delay between each communication link in the current network.
Step S302, inputting the network state information into a target flow scheduling measurement generation model to obtain a preset flow scheduling strategy.
Step S303, carrying out flow scheduling based on a preset flow scheduling strategy.
Therefore, the flow scheduling method provided by the embodiment of the invention can timely and comprehensively perform flow scheduling by the software defined network aiming at the current state of the network in the Data Center (Data Center), so that the problems of high network communication delay and network congestion possibly existing are avoided.
In addition, it should be further noted that, in the above embodiments of the method, the steps of the various methods are divided, for clarity of description, and may be combined into one step or split into multiple steps when implemented, so long as they include the same logic relationship, and they are all within the protection scope of the present patent; it is within the scope of this patent to add insignificant modifications to the algorithm or flow or introduce insignificant designs, but not to alter the core design of its algorithm and flow.
The embodiment of the invention also provides a training device of the flow scheduling strategy generation model, as shown in fig. 4, the training device 400 of the flow scheduling strategy generation model includes: an information acquisition module 401, a link delay prediction model acquisition module 402, a generation module 403, and a training module 403.
The information obtaining module 401 is configured to obtain traffic information of each network node, and obtain link delay information of a communication link between each network node.
The link delay prediction model obtaining module 402 is configured to obtain a target link delay prediction model according to the traffic information and the link delay information.
The generating module 403 is configured to generate training data according to preset service quality information, a preset traffic matrix, and a target link delay prediction model, where the preset service quality information is used to represent service quality information of a preset service, and the training data represents traffic scheduling policies for the preset service in different network states.
The training module 404 is configured to train the initial traffic scheduling policy generation model according to training data to obtain a target traffic scheduling policy generation model, where the target traffic scheduling measurement generation model is configured to generate traffic scheduling policies for each network node.
In some embodiments, the link delay prediction model acquisition module 402, when acquiring the target link delay prediction model according to the traffic information and the link delay information, may be configured to: generating a plurality of communication links according to the flow information and the link delay information, wherein the communication links are formed by a first network node and a second network node, and the first network node and the second network node are any different network nodes in all the network nodes; constructing a communication link diagram according to a plurality of communication links, wherein each network node in the communication link diagram takes each network node as a diagram node, and at least traffic information and link delay between different communication links are taken as connecting edges between the corresponding diagram nodes; training an initial link delay prediction model by using a communication link diagram to obtain a target link delay prediction model, wherein the target link delay prediction model is used for predicting link delay between each network node.
In some embodiments, the generating module 403 may be configured to, when generating training data according to the preset qos information, the preset traffic matrix, and the target link delay prediction model: acquiring a target communication path set among all network nodes according to preset service quality information, a preset traffic matrix and a target link delay prediction model, wherein the target communication path set consists of communication paths meeting preset conditions in communication paths for forwarding traffic of preset services under the network state represented by the preset traffic matrix of all network nodes; and generating training data according to the preset service quality information, the preset traffic matrix and the target communication path set.
In some embodiments, the generating module 403, when obtaining the target communication path set between the network nodes according to the preset service quality information, the preset traffic matrix and the target link delay prediction model, may be configured to: generating a plurality of flow matrix combinations according to the flow type of the preset service and the preset flow matrix corresponding to the service quality information; and calculating communication paths which correspond to the plurality of traffic matrix combinations and meet preset conditions by using a preset routing algorithm so as to obtain a target communication path set.
In some embodiments, the training module 404, when training the initial traffic scheduling policy generation model according to training data, may be configured to: acquiring an initial flow scheduling strategy generation model constructed based on a BP neural network model structure; inputting training data into an initial flow scheduling measurement model to obtain a predicted flow scheduling strategy; calculating an error value between a predicted link delay corresponding to the predicted traffic scheduling policy and an expected link delay corresponding to the communication path in the training data; and adjusting parameters of the initial flow scheduling strategy by taking the error value as a loss value to obtain an initial flow scheduling strategy generation model meeting preset convergence conditions as a target flow scheduling strategy generation model.
In some embodiments, the information obtaining module 401, when obtaining traffic information of each network node, may be configured to: analyzing communication data among all network nodes by using a preset network flow analysis tool to obtain initial flow information of all network nodes; classifying the initial flow information according to the service type corresponding to the initial flow information; and filtering and normalizing the classified initial flow information to obtain flow information.
The embodiment of the present invention further provides a traffic scheduling device, as shown in fig. 5, where the traffic scheduling device 500 includes: a status information acquisition module 501, a traffic scheduling policy acquisition module 502, and a traffic scheduling module 503.
The state information obtaining module 501 is configured to obtain network state information of a current network, where the network state information includes at least a service type of a service to be currently traffic scheduled and a predicted link delay between communication links in the current network.
The flow scheduling policy obtaining module 502 is configured to input network status information into a target flow scheduling measurement generation model, and obtain a preset flow scheduling policy.
The traffic scheduling module 503 is configured to perform traffic scheduling based on a preset traffic scheduling policy.
The functions or modules included in the apparatus provided by the embodiments of the present invention may be used to perform the methods described in the corresponding method embodiments, and the specific implementation and technical effects of the methods may refer to the descriptions of the method embodiments above, which are not repeated herein for brevity.
In this embodiment, each module is a logic module, and in practical application, one logic unit may be one physical unit, or may be a part of one physical unit, or may be implemented by a combination of a plurality of physical units. In addition, in order to highlight the innovative part of the present invention, units that are not so close to solving the technical problem presented by the present invention are not introduced in the present embodiment, but this does not indicate that other units are not present in the present embodiment.
Referring to fig. 6, an embodiment of the present invention provides an electronic device including:
one or more processors 601;
a memory 602 having one or more programs stored thereon, which when executed by one or more processors, cause the one or more processors to implement the training method or the traffic scheduling method of the traffic scheduling policy generation model of any of the above embodiments;
One or more I/O interfaces 603, coupled between the processor and the memory, are configured to enable information interaction of the processor with the memory.
Wherein the processor 601 is a device having data processing capabilities including, but not limited to, a Central Processing Unit (CPU) or the like; memory 602 is a device with data storage capability including, but not limited to, random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically charged erasable programmable read-only memory (EEPROM), FLASH memory (FLASH); an I/O interface (read/write interface) 603 is connected between the processor 601 and the memory 602, and enables information interaction between the processor 601 and the memory 602, including but not limited to a data Bus (Bus) or the like.
In some embodiments, processor 601, memory 602, and I/O interface 603 are interconnected by a bus to further connect with other components of a computing device.
The embodiment of the present invention further provides a computer readable medium, on which a computer program is stored, where the program when executed by a processor implements the training method or the traffic scheduling method of the traffic scheduling policy generation model in any one of the foregoing embodiments, and specific steps are not repeated herein to avoid repetitive description.
Those of ordinary skill in the art will appreciate that all or some of the steps, systems, functional modules/units in the apparatus, and methods of the invention described above may be implemented as software, firmware, hardware, and suitable combinations thereof. In a hardware implementation, the division between the functional modules/units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be performed cooperatively by several physical components. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). The term computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as known to those skilled in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by a computer. Furthermore, as is well known to those of ordinary skill in the art, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
Those skilled in the art will appreciate that while some embodiments described herein include some features but not others included in other embodiments, combinations of features of different embodiments are meant to be within the scope of embodiments of the invention and form different embodiments.
It is to be understood that the above embodiments are merely illustrative of the application of the principles of the present invention, but not in limitation thereof. Various modifications and improvements may be made by those skilled in the art without departing from the spirit and substance of the invention, and are also considered to be within the scope of the invention.
Claims (10)
1. The training method of the traffic scheduling strategy generation model is characterized by comprising the following steps of:
acquiring flow information of each network node and link delay information of communication links between the network nodes;
acquiring a target link delay prediction model according to the flow information and the link delay information;
generating training data according to preset service quality information, a preset flow matrix and the target link delay prediction model, wherein the preset service quality information is used for representing service quality information of preset service, and the training data represents flow scheduling strategies aiming at the preset service in different network states;
training an initial flow scheduling strategy generation model according to the training data to obtain a target flow scheduling strategy generation model, wherein the target flow scheduling measurement generation model is used for generating flow scheduling strategies for all network nodes.
2. The method of claim 1, wherein the obtaining a target link delay prediction model from the traffic information and the link delay information comprises:
generating a plurality of communication links according to the traffic information and the link delay information, wherein the communication links are formed by a first network node and a second network node, and the first network node and the second network node are any different network nodes in the network nodes;
Constructing a communication link diagram according to the plurality of communication links, wherein the communication link diagram takes each network node in the plurality of communication links as a diagram node, and takes at least traffic information and link delay among different communication links as connecting edges among the corresponding diagram nodes;
training an initial link delay prediction model by using the communication link graph to obtain the target link delay prediction model, wherein the target link delay prediction model is used for predicting link delay between all network nodes.
3. The method of claim 2, wherein generating training data based on the preset quality of service information, the preset traffic matrix, and the target link delay prediction model comprises:
acquiring a target communication path set among the network nodes according to the preset service quality information, the preset traffic matrix and the target link delay prediction model, wherein the target communication path set consists of communication paths which meet preset conditions in communication paths for forwarding traffic of the preset service by the network nodes in a network state represented by the preset traffic matrix;
And generating the training data according to the preset service quality information, the preset flow matrix and the target communication path set.
4. The method of claim 3, wherein the obtaining the set of target communication paths between the network nodes based on the preset quality of service information, the preset traffic matrix, and the target link delay prediction model comprises:
generating a plurality of flow matrix combinations according to the flow type of the preset service corresponding to the service quality information and the preset flow matrix;
and calculating communication paths which correspond to the plurality of traffic matrix combinations and meet the preset conditions by using a preset routing algorithm so as to obtain the target communication path set.
5. The method of claim 3, wherein training the initial traffic scheduling policy generation model according to the training data to obtain the target traffic scheduling policy generation model comprises:
acquiring the initial flow scheduling strategy generation model constructed based on the BP neural network model structure;
inputting the training data into the initial flow scheduling measurement model to obtain a predicted flow scheduling strategy;
Calculating an error value between a predicted link delay corresponding to the predicted traffic scheduling policy and an expected link delay corresponding to a communication path in the training data;
and adjusting the parameters of the initial flow scheduling strategy by taking the error value as a loss value to obtain an initial flow scheduling strategy generation model meeting a preset convergence condition as the target flow scheduling strategy generation model.
6. The method according to claim 1, wherein the obtaining traffic information of each network node comprises:
analyzing the communication data among the network nodes by using a preset network flow analysis tool to obtain initial flow information of the network nodes;
classifying the initial flow information according to the service type corresponding to the initial flow information;
and filtering and normalizing the initial flow information after the classification processing to obtain the flow information.
7. A traffic scheduling method, comprising:
acquiring network state information of a current network, wherein the network state information at least comprises a service type of a service to be subjected to flow scheduling currently and a predicted link delay between communication links in the current network;
Inputting the network state information into a target flow scheduling measurement generation model to obtain a preset flow scheduling strategy;
and carrying out flow scheduling based on the preset flow scheduling strategy.
8. A training device for a traffic scheduling policy generation model, comprising:
the information acquisition module is used for acquiring the flow information of each network node and acquiring the link delay information of the communication links between the network nodes;
the link delay prediction model acquisition module is used for acquiring a target link delay prediction model according to the flow information and the link delay information;
the generation module is used for generating training data according to preset service quality information, a preset flow matrix and the target link delay prediction model, wherein the preset service quality information is used for representing service quality information of preset service, and the training data represents flow scheduling strategies aiming at the preset service in different network states;
the training module is used for training the initial flow scheduling strategy generation model according to the training data to obtain a target flow scheduling strategy generation model, wherein the target flow scheduling measurement generation model is used for generating flow scheduling strategies for all the network nodes.
9. An electronic device, comprising:
one or more processors;
a memory having one or more programs stored thereon, which when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-7;
one or more I/O interfaces coupled between the processor and the memory configured to enable information interaction of the processor with the memory.
10. A computer readable medium having stored thereon a computer program which, when executed by a processor, implements the method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310032821.2A CN116055378B (en) | 2023-01-10 | 2023-01-10 | Training method and device for traffic scheduling strategy generation model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310032821.2A CN116055378B (en) | 2023-01-10 | 2023-01-10 | Training method and device for traffic scheduling strategy generation model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116055378A true CN116055378A (en) | 2023-05-02 |
| CN116055378B CN116055378B (en) | 2024-05-28 |
Family
ID=86112878
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310032821.2A Active CN116055378B (en) | 2023-01-10 | 2023-01-10 | Training method and device for traffic scheduling strategy generation model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116055378B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090180480A1 (en) * | 2008-01-15 | 2009-07-16 | Lucent Technologies Inc. | Systems and methods for network routing |
| WO2016123040A1 (en) * | 2015-01-30 | 2016-08-04 | Hewlett-Packard Development Company, L.P. | Adjusted spanning tree protocol path cost values in a software defined network |
| CN111245718A (en) * | 2019-12-30 | 2020-06-05 | 浙江工商大学 | Routing optimization method based on SDN context awareness |
| CN113114573A (en) * | 2021-04-16 | 2021-07-13 | 重庆邮电大学 | Video stream classification and scheduling system in SDN network |
| CN113158543A (en) * | 2021-02-02 | 2021-07-23 | 浙江工商大学 | Intelligent prediction method for software defined network performance |
| CN113489654A (en) * | 2021-07-06 | 2021-10-08 | 国网信息通信产业集团有限公司 | Routing method, routing device, electronic equipment and storage medium |
| CN113778648A (en) * | 2021-08-31 | 2021-12-10 | 重庆理工大学 | Task scheduling method based on deep reinforcement learning in hierarchical edge computing environment |
| CN113938415A (en) * | 2021-09-09 | 2022-01-14 | 华中科技大学 | Network route forwarding method and system based on link state estimation |
| CN114827021A (en) * | 2022-06-27 | 2022-07-29 | 南京邮电大学 | Multimedia service flow acceleration system based on SDN and machine learning |
| CN115225561A (en) * | 2022-08-15 | 2022-10-21 | 南京邮电大学 | Route optimization method and system based on graph structure characteristics |
| CN115499306A (en) * | 2022-07-29 | 2022-12-20 | 天翼云科技有限公司 | Method and device for constructing traffic scheduling model, electronic equipment and storage medium |
-
2023
- 2023-01-10 CN CN202310032821.2A patent/CN116055378B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090180480A1 (en) * | 2008-01-15 | 2009-07-16 | Lucent Technologies Inc. | Systems and methods for network routing |
| WO2016123040A1 (en) * | 2015-01-30 | 2016-08-04 | Hewlett-Packard Development Company, L.P. | Adjusted spanning tree protocol path cost values in a software defined network |
| CN111245718A (en) * | 2019-12-30 | 2020-06-05 | 浙江工商大学 | Routing optimization method based on SDN context awareness |
| CN113158543A (en) * | 2021-02-02 | 2021-07-23 | 浙江工商大学 | Intelligent prediction method for software defined network performance |
| CN113114573A (en) * | 2021-04-16 | 2021-07-13 | 重庆邮电大学 | Video stream classification and scheduling system in SDN network |
| CN113489654A (en) * | 2021-07-06 | 2021-10-08 | 国网信息通信产业集团有限公司 | Routing method, routing device, electronic equipment and storage medium |
| CN113778648A (en) * | 2021-08-31 | 2021-12-10 | 重庆理工大学 | Task scheduling method based on deep reinforcement learning in hierarchical edge computing environment |
| CN113938415A (en) * | 2021-09-09 | 2022-01-14 | 华中科技大学 | Network route forwarding method and system based on link state estimation |
| CN114827021A (en) * | 2022-06-27 | 2022-07-29 | 南京邮电大学 | Multimedia service flow acceleration system based on SDN and machine learning |
| CN115499306A (en) * | 2022-07-29 | 2022-12-20 | 天翼云科技有限公司 | Method and device for constructing traffic scheduling model, electronic equipment and storage medium |
| CN115225561A (en) * | 2022-08-15 | 2022-10-21 | 南京邮电大学 | Route optimization method and system based on graph structure characteristics |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116055378B (en) | 2024-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3932025B1 (en) | Computing resource scheduling method, scheduler, internet of things system, and computer readable medium | |
| CN105379204B (en) | Method and system for the resource for selecting data to route | |
| US11296963B1 (en) | System and method of traffic prediction for IoT nodes | |
| CN112235389A (en) | Intelligent terminal data processing method and device based on Internet of things | |
| CN110297713A (en) | Configuration management system and method of cloud host | |
| US10057857B2 (en) | System power management and optimization in telecommunication systems | |
| US20150281008A1 (en) | Automatic derivation of system performance metric thresholds | |
| CN116055378B (en) | Training method and device for traffic scheduling strategy generation model | |
| CN114595970A (en) | Resource scheduling intelligent decision method and device, electronic equipment and storage medium | |
| CN116055406B (en) | Training method and device for congestion window prediction model | |
| CN116521344B (en) | AI algorithm scheduling method and system based on resource bus | |
| CN112534399A (en) | Semantic-based Internet of things equipment data processing related application program installation method and device | |
| CN112085282B (en) | Cell traffic prediction method and server | |
| US9047572B2 (en) | Mode determination for multivariate time series data | |
| CN111756798B (en) | Service scheduling method, device, equipment and storage medium based on gateway cascade | |
| CN113487344B (en) | Communication service prediction method, centralized server, edge service terminal and medium | |
| CN114629835B (en) | Route configuration method, device, computer readable storage medium and electronic equipment | |
| US20230221934A1 (en) | Application deployment method, application blueprint generation method, platform, terminal and medium | |
| WO2024093219A1 (en) | Service path optimization method and apparatus, and storage medium | |
| CN116781539A (en) | AI capability scheduling method and device, electronic equipment and storage medium | |
| CN110309045B (en) | Method, apparatus, medium and computing device for determining future state of server | |
| CN116489091A (en) | Flow scheduling method and device based on remote in-band telemetry and time delay | |
| CN115037625A (en) | Network slice processing method and device, electronic equipment and readable storage medium | |
| CN116471271A (en) | Unmanned gas station data interaction state adjusting method, equipment and storage medium | |
| CN113746899A (en) | Cloud platform access method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |