CN111371644B - Multi-domain SDN network traffic situation prediction method and system based on GRU - Google Patents
Multi-domain SDN network traffic situation prediction method and system based on GRU Download PDFInfo
- Publication number
- CN111371644B CN111371644B CN202010129348.6A CN202010129348A CN111371644B CN 111371644 B CN111371644 B CN 111371644B CN 202010129348 A CN202010129348 A CN 202010129348A CN 111371644 B CN111371644 B CN 111371644B
- Authority
- CN
- China
- Prior art keywords
- situation
- gru
- domain
- network
- flow
- 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.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
- H04L43/0876—Network utilisation, e.g. volume of load or congestion level
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/147—Network analysis or design for predicting network behaviour
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Environmental & Geological Engineering (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a GRU-based multi-domain SDN network traffic situation prediction method and a GRU-based multi-domain SDN network traffic situation prediction system, which comprise the following steps: collecting historical flow situation values, and constructing a sample set in a sliding window mode; carrying out data normalization on the sample data, and cutting the sample data into a training set and a test set; establishing a GRU model, and optimizing GRU model parameters by adopting an SSA algorithm; wherein the GRU model is configured to: extracting key elements influencing network flow from the historical flow situation value; then, carrying out weighted quantitative calculation on the flow situation value according to related factors, and evaluating the current flow state; and (4) performing situation prediction on the test set by using the trained SSA-GRU neural network. The SSA algorithm is used for optimizing the weight of the GRU algorithm, the optimized searching direction is effectively controlled, and the convergence efficiency can be better improved.
Description
Technical Field
The invention belongs to the technical field of network traffic, and particularly relates to a method and a system for predicting the situation of multi-domain SDN network traffic based on GRUs.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
With the continuous development of cloud computing, virtualization and other related technologies, the network application requirements are more and more complex, the problems that the traditional network is difficult to expand, the configuration complexity is high and the like also occur, and various virtualization technologies need to rapidly and flexibly allocate network resources, so that the traditional network system architecture is more difficult to meet the requirements. Against the background of such demands, software Defined Networking (SDN) has emerged. Compared with a traditional network, the SDN is mainly characterized by:
(1) Centrally controlled network architecture
The centralized control is expressed in that the SDN can send commands to the network through a programmable controller, and the flow table items on the OpenFlow switches are controlled to control the storage, forwarding or discarding of data packets on each switch, so that the controller can realize omnibearing control on the underlying network.
(2) Open network API (Application Programming Interface)
The programmable design implemented by the SDN controller enables the controller to implement development and design of most network management functions, and user-oriented services are generally abstracted into application layer functions. When the management operations such as function upgrading and reconfiguration are performed on the SDN, each bottom layer hardware device, namely each switch or router, does not need to be configured independently like a traditional network architecture, only secondary development is performed on an application layer, and complexity of network management and function deployment is reduced.
(3) Implementing network virtualization
Network virtualization enables a network to achieve greater network resource utilization. The virtualization of the network function is not only for a single entity in the network, but also for the network as a whole, so that the whole network becomes a device which can be flexibly configured and adjusted in real time. The SDN controller enables the SDN network not to be limited to a certain single network by abstracting a hardware platform and a software control center, so that multi-network fusion is better realized.
Analysis of network traffic has been the focus of SDN research. With the continuous development of the multi-domain SDN, frequent information interaction causes the network traffic to increase dramatically, increasing the complexity of the network operating condition. Because the network traffic situation prediction has the characteristics of complexity, nonlinearity, time sequence and the like, the traditional prediction method cannot accurately and rapidly predict the network situation in a future period of time according to the collected historical network traffic situation information, and a unified prediction model is difficult to establish.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a GRU-based multi-domain SDN network flow situation prediction method, which carries out planning decision on a multi-domain SDN network in advance and maintains the network stability to a greater extent.
In order to achieve the above object, one or more embodiments of the present invention provide the following technical solutions:
a multi-domain SDN network traffic situation prediction method based on GRU comprises the following steps:
collecting historical flow situation values, and constructing a sample set in a sliding window mode;
carrying out data normalization on sample data, and cutting the sample data into a training set and a test set;
establishing a GRU model, and optimizing GRU model parameters by adopting an SSA algorithm;
wherein the GRU model is configured to:
extracting key elements influencing network flow from historical flow situation values; then, carrying out weighting quantitative calculation on the flow situation value according to related factors, and evaluating the current flow state;
and (4) performing situation prediction on the test set by using the trained SSA-GRU neural network.
According to the technical scheme, the GRU model is composed of an SDN domain, a database, a situation evaluation server and a situation prediction server, the SDN domain adopts a vertical framework, the database stores flow situation element values, situation values and prediction value data, and the situation evaluation server and the situation prediction server are used for evaluating the network flow situation in the current time period and predicting the network flow situation in the next time period.
According to a further technical scheme, the flow situation elements are extracted: the system comprises a controller, a database and a plurality of switches, wherein the controller is connected with each switch so as to obtain a large amount of real-time and historical flow data, analyze and process the collected original data, extract key elements influencing the network flow state, form a uniform data format and store the key elements in the database;
the multi-domain SDN network adopts a vertical architecture and comprises inter-domain controllers and intra-domain controllers, intra-domain data are collected by the intra-domain controllers, and inter-domain data are collected by the inter-domain controllers.
According to the further technical scheme, the situation assessment server obtains a numerical value of the network traffic situation in the current time period by carrying out weighted quantitative calculation on the traffic situation element values in the database, writes the current time period situation values into the database and transmits the current time period situation values to the situation prediction server, so that the current traffic state of the multi-domain SDN network is quantitatively described, and the current network traffic situation is determined.
According to the further technical scheme, a situation prediction server predicts the change trend of network flow by using historical and current multi-domain SDN network flow situation values, wherein the situation values at the historical moment are provided by a database, the situation values at the current moment are provided by a situation evaluation server, and the predicted values are written into the database and transmitted into an inter-domain controller;
the network flow situation prediction adopts a method of combining an SSA algorithm and a GRU algorithm, namely, the SSA algorithm is introduced on the basis of the advantages and the characteristics of the GRU.
According to the technical scheme, the flow situation level of the next period is divided into a plurality of classes according to the corresponding standard of the predicted flow situation value and the reference flow situation value;
and carrying out planning decision on the multi-domain SDN network in advance according to the predicted traffic situation level of the multi-domain SDN network.
In the further technical scheme, the SSA algorithm is used for optimizing the weight of the GRU algorithm and controlling the optimized search direction.
GRU-based multi-domain SDN network traffic situation prediction system comprises: a GRU model configured to:
extracting key elements influencing network flow from historical flow situation values; then, carrying out weighting quantitative calculation on the flow situation value according to related factors, and evaluating the current flow state;
carrying out situation prediction on the test set by using the trained SSA-GRU neural network;
the prediction model comprises an SDN domain, a database, a situation evaluation server and a situation prediction server, wherein the SDN domain adopts a vertical architecture, the database stores flow situation element values, situation values and prediction value data, and the situation evaluation server and the situation prediction server are used for evaluating the current period network flow situation and predicting the next period network flow situation.
According to the further technical scheme, the situation assessment server obtains a numerical value of the network traffic situation in the current time period by carrying out weighted quantitative calculation on the traffic situation element values in the database, writes the current time period situation values into the database and transmits the current time period situation values to the situation prediction server, so that the current traffic state of the multi-domain SDN network is quantitatively described, and the current network traffic situation is determined.
According to the further technical scheme, a situation prediction server predicts the change trend of network traffic by using historical and current multi-domain SDN network traffic situation values, wherein the historical moment situation values are provided by a database, the current moment situation values are provided by a situation evaluation server, and predicted values are written into the database and transmitted into an inter-domain controller;
the network flow situation prediction adopts a method of combining an SSA algorithm and a GRU algorithm, namely, the SSA algorithm is introduced on the basis of the advantages and the characteristics of the GRU.
The above one or more technical solutions have the following beneficial effects:
the invention can provide more possibilities for improving the accuracy of the network traffic situation prediction by searching and trying more novel prediction models and exploring and optimizing more prediction algorithms.
According to the invention, the SSA algorithm is used for optimizing the weight of the GRU algorithm, so that the optimized searching direction is effectively controlled, and the convergence efficiency can be better improved.
The method and the system perform planning decision on the multi-domain SDN network in advance aiming at the predicted traffic situation level of the multi-domain SDN network, and maintain the stability of the network to a greater extent.
The method is combined with the characteristics of the multi-domain SDN network to design a network flow situation prediction model. The prediction model is composed of an SDN domain, a database, a situation evaluation server and a situation prediction server. The multi-domain SDN network adopts a vertical architecture, a database is introduced to store data such as flow situation element values, situation values and predicted values, and meanwhile, a network situation evaluation server and a situation prediction server are introduced to evaluate the network flow situation in the current time period and predict the network flow situation in the next time period.
The invention provides a multi-domain SDN network traffic situation prediction algorithm based on SSA-GRU. The SSA-GRU algorithm is adopted for predicting the network traffic situation, namely, the SSA algorithm is introduced on the basis of the advantage characteristics of the GRU. The traditional GRU algorithm optimizes parameters through the BPTT algorithm, but the BPTT algorithm has the defects of high complexity, easiness in convergence to local optimum and the like, and the SSA algorithm is adopted for optimizing the weight of the GRU algorithm, so that the optimized search direction is effectively controlled, and the convergence efficiency can be better improved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
FIG. 1 is a flow chart of network traffic situation prediction according to an embodiment of the present invention;
fig. 2 is a model diagram of multi-domain SDN network traffic situation prediction according to an embodiment of the present invention;
FIG. 3 is a diagram of a gated loop unit network according to an embodiment of the present invention;
FIG. 4 is a diagram of a GRU unit structure according to an embodiment of the present invention;
FIG. 5 is a flow chart of flow situation prediction based on the SSA-GRU algorithm in an embodiment of the present invention.
Detailed Description
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The embodiments and features of the embodiments of the invention may be combined with each other without conflict.
The overall concept of the present disclosure is:
extracting key elements influencing network flow;
evaluating the flow situation: weighting the key elements to calculate the current flow situation value and evaluating the current flow state;
predicting the flow situation: predicting a multi-domain SDN network flow situation value in the next time period by using a trained SSA-GRU algorithm according to the historical flow situation value and the current flow situation value;
planning and deciding a multi-domain SDN in advance according to situation levels corresponding to predicted flow situation values
Based on this, the network traffic situation prediction model established in the multi-domain SDN comprises the following steps: the system comprises an SDN domain, a database, a situation assessment server and a situation prediction server.
A database: acquiring current time interval flow data, extracting key elements of the flow data, and writing the key elements into a database;
situation assessment server: and carrying out weighted quantitative calculation on key elements of the flow situation in the database. Obtaining a numerical value of the current time period flow situation, writing the current time period situation value into a database and transmitting the current time period situation value to a situation prediction server;
situation prediction server: and predicting the situation value at the next moment by using the trained SSA-GRU algorithm by using the historical situation value stored in the database and the situation value provided by the prediction server.
Example one
The embodiment discloses a GRU-based multi-domain SDN network traffic situation prediction method, which comprises the following steps: 1) Collecting historical flow situation values (obtained from a database part in a prediction model), and constructing a sample set in a sliding window mode; 2) And carrying out data normalization on the sample data, and cutting the sample data into a training set and a test set. The training set is used for training a GRU model, and the testing set is used for verifying the effectiveness of the trained model; 3) Establishing a GRU model, and optimizing GRU model parameters by adopting an SSA algorithm; 4) And (4) carrying out situation prediction on the test set by using the trained SSA-GRU neural network, and verifying the training effect.
In order to further enable a network manager to predict the network traffic variation trend in advance and provide decision support for the network manager to deploy and regulate the network in time, great necessity exists for prediction research on the multi-domain SDN network traffic situation. As shown in fig. 1, a situation prediction of multi-domain SDN network traffic is performed, and first, key elements affecting network traffic are extracted; then, carrying out weighted quantitative calculation on the flow situation value according to related factors, and evaluating the current flow state; secondly, predicting the network traffic situation according to history and current information, and predicting a multi-domain SDN network traffic situation value in the next time period by using a trained SSA-GRU algorithm; and finally, planning and deciding the multi-domain SDN in advance according to the situation level corresponding to the predicted flow situation value. Based on this, a multi-domain SDN network traffic situation prediction model is provided, which is composed of an SDN domain, a database, a situation evaluation server and a situation prediction server, as shown in FIG. 2.
(1) Extracting flow situation elements
The SDN has the characteristic of separation of a control layer and a forwarding layer, is connected with each switch through a controller, so that a large amount of real-time and historical flow data are obtained, the collected original data are analyzed and processed, key elements (throughput, packet loss rate, maximum transmission delay, switch data Packet sending rate and receiving rate, packet _ in rate, cross-domain request rate, REST request rate and the like) influencing the flow state of the network are extracted, and a uniform data format is formed and stored in a database so as to facilitate the next processing. The multi-domain SDN network adopts a vertical architecture and comprises inter-domain controllers and intra-domain controllers, intra-domain data are collected by the intra-domain controllers, and inter-domain data are collected by the inter-domain controllers.
(2) Evaluating flow situation
The situation evaluation server obtains a numerical value of the network traffic situation in the current time period by carrying out weighted quantitative calculation on the traffic situation element values in the database, writes the current time period situation value into the database and transmits the current time period situation value to the situation prediction server, and realizes quantitative description on the current traffic state of the multi-domain SDN network, so that the current network traffic situation is determined, and the network traffic situation is used for quantitatively describing the current network traffic state.
(3) Predicting flow situation
And the situation prediction server predicts the change trend of the network traffic by using the historical and current multi-domain SDN network traffic situation values, wherein the historical moment situation values are provided by a database, the current moment situation values are provided by the situation evaluation server, and the predicted values are written into the database and are transmitted into the inter-domain controller. The network flow situation prediction adopts a method of combining an SSA algorithm and a GRU algorithm, namely, the SSA algorithm is introduced on the basis of the advantage characteristics of the GRU. The traditional GRU algorithm optimizes parameters through the BPTT algorithm, but the BPTT algorithm has the defects of high complexity, easiness in convergence to local optimum and the like, and the SSA algorithm is used for optimizing the weight of the GRU algorithm, so that the optimized search direction is effectively controlled, and the convergence efficiency can be better improved.
(4) Assessing network traffic situation levels
And classifying the flow situation levels of the next period into 4 classes according to the corresponding standard of the predicted flow situation values and referring to the flow situation values in the table 1. Aiming at the predicted traffic situation level of the multi-domain SDN network, an administrator makes planning decisions on the multi-domain SDN network in advance, and the stability of the network is maintained to a great extent.
TABLE 1 Multi-domain SDN network traffic situation level table
The multi-domain SDN network traffic situation prediction method related to the specific example of the present application is described in detail as follows:
GRU neural network:
the GRU is adopted aiming at the part of predicting the flow situation, and the GRU is a gated cyclic neural network and is good at processing time sequence data. The structure diagram of the GRU network consists of three parts, namely an input layer, a hidden layer and an output layer, as shown in fig. 3.
The hidden layer is composed of GRU units, and the structure of the GRU unit is shown in fig. 4.
ht-1 is the state at the time of t-1; x is the number of t And h t The input and the output of the GRU unit at the time t are respectively;
is in a hidden state; z is a radical of t And r are the update gate and the reset gate, respectively. For one input, a sequence of time length T x = (x) 1 ,x 2 ,...,x t ,...,x T ) The GRU performs forward propagation calculation from T =1 to T in time series. First, the current time is input with x t And the state h of the previous moment t-1 A compute update gate and a reset gate.
Updating the door z t The method is used for solving the gradient attenuation problem in the recurrent neural network, determines the degree of selecting the information of the previous moment at the current moment, and has the expression formulas (1) and (2).
Wherein sigma is a sigmoid function; w z And U z Weights of the input and last hidden layer to the refresh gate are respectively;
the closer to 1, the more information of the previous time is selected as the current time information.
Reset gate r t The method is used for abandoning the past hidden state irrelevant to the future and controlling the influence degree of the current moment by the information of the previous moment, and the expressions are shown in formulas (3) and (4).
Wherein the content of the first and second substances, Wr and Ur respectively input and last oneWeights for etching the hidden layer to the reset gate; r t closer to 0, the current time is less affected by the previous time.
After the update gate and the reset gate are calculated, the hidden state at the current moment is updated through equations (5) and (6), and the output of the hidden layer GRU unit is obtained as shown in equation (7).
Wherein, the first and the second end of the pipe are connected with each other,
and &>
Weights of the hidden layer to the hidden state at the input and the previous moment are respectively; . Is a Hadamard product, i.e. the product of the corresponding elements.
And finally, sending the output of the hidden layer to the full-connection layer to obtain the output of the output layer as shown in formulas (8) and (9).
Wherein, W o Is the weight of the hidden layer to the output layer.
The sea squirt group algorithm (Salp Swarm algorithm, SSA) is a novel group intelligent optimization algorithm, different from other algorithms, and is not distributed in a 'group' manner, but adopts a head-to-tail connection manner to form a 'chain' form, and the 'chain' form is followed by movement in sequence. Each individual goblet ascidian represents a parameter configuration of a GRU neural network, a leader in a goblet ascidian chain is arranged at the forefront of a team and has the optimal judgment on the environment, but different from other group intelligent algorithms, the leader does not directly influence the moving direction of the whole population but directly influences the position update of the next second individual goblet ascidian, the second individual influences the third individual, and the like. Therefore, the influence degree of the position of the leader on the positions of the rest goblet and sea squirts is reduced in a proper way, so that the individuals who are ranked behind have better diversity.
In the algorithm of goblet sea squirt group, the predation space is an Euclidean space with dimension NxD, where N is the scale of goblet sea squirt group and D is the space dimension. Presence of food in space F = [ F = [) 1 F 2 … F D ] T The location of the nth goblet sea squirt can be represented as X n ＝[X n1 X n2 … X nD ] T N =1,2, \ 8230;, N. The upper bound of the search space is expressed as ub = [ ub = 1 ub 2 … ub D ]Lower bound is lb = [ lb ] 1 lb 2 … lb D ]. Randomly initializing a population:
X N×D ＝rand(N,D)×(ub-lb)+lb (10)
in the population, the state of each dimension of the leader is
Where D =1,2, \8230;, D denotes the dimension of the leader, m =2,3, \8230;, N denotes the serial number of the follower.
The leader is responsible for searching for food in the environment and guiding the movement of the whole group, so that the position update of the leader has strong randomness, and the update follows the formula
The control parameter in the formula is only c 1 ，c 2 And c 3 3 in which c 2 And c 3 Are all [0,1 ]]The random numbers between the leader and the leader are used for enhancing the randomness of the movement of the leader and enhancing the global searching capability and the diversity of individuals. c. C 1 Is called convergence factor when c 1 >1, performing global search; when c is 1 <1, an accurate estimation value is obtained for the local development. Convergence factor c taken by SSA 1 The expression is shown in the formula.
Wherein l is the current iteration number, l max Is the maximum number of iterations, c 1 Is a decreasing function with the value range of 0-2.
In SSA, the follower does not have random motion, but follows the motion in a chain shape in sequence, so the position of the follower is only related to the initial position, the motion speed and the acceleration in the motion process, the motion mode conforms to Newton motion law, and the motion distance R of the follower can be expressed as
In the optimization iteration process, the time t is the difference of the iteration times, so t =1; v. of 0 Is the initial velocity of the follower, and the velocity of the follower is 0 at the beginning of each iteration; a is the acceleration of the follower between the beginning of an iteration and the end of the iteration, and the calculation formula is a = (v) final -v 0 ) T, since the follower only follows the sheath of the previous bottle, the speed of movement is high
Wherein t =1,v is known 0 =0, and thus, formula (13) may be expressed as
Follower updates follow the equation
Wherein the content of the first and second substances,
is the d-dimensional position of the mth follower before the update>
Is the updated follower position.
And (3) traffic situation prediction based on the SSA-GRU neural network:
the traditional GRU algorithm optimizes parameters through a BPTT algorithm, but the BPTT algorithm has the defects of high complexity, easiness in convergence to local optimization and the like. According to the method, the SSA algorithm is used for optimizing the weight of the GRU algorithm, so that the optimized search direction is effectively controlled, and the convergence efficiency can be better improved; the situation prediction of the SSA-GRU neural network is mainly divided into three parts:
data preprocessing, SSA optimization of GRU neural network parameters, and prediction output of the GRU neural network. According to the flow chart shown in fig. 5, the detailed flow of the traffic situation prediction based on the SSA-GRU algorithm is as follows:
step 1: and collecting historical flow situation values, and constructing a sample set in a sliding window mode.
Step 2: and carrying out data normalization on the sample data, and cutting the sample data into a training set and a test set. The training set is used for training an initial GRU model, and the testing set is used for verifying the effectiveness of the trained model.
And step 3: and establishing a GRU model.
And 4, step 4: initializing the maximum iteration times L, the size N of the goblet sea squirt population and the position D. The population position D is initialized using equation (10).
The following steps are used for training a GRU model, and obtaining the optimal individual, namely the optimal network parameter value through SSA
And 5: calculating goblet sea squirtIndividual fitness values in the population. For goblet sea squirt individual, its position X n ＝[X n1 X n2 …X nD ] T Decomposing into corresponding GRU network parameters, and calculating fitness function of N goblet ascidians
Wherein y' is the output value of the LSTM model, y is the true value, and when the fitness reaches the minimum, the LSTM model is the optimal LSTM model.
Step 6: and selecting the food. Sorting the goblet sea squirt groups according to the fitness value, and setting the position of the goblet sea squirt with the first fitness as the current food position.
And 7: and selecting the leader and the follower. After the food is selected, the N-1 goblet ascidians in the group are selected, the goblet ascidians with the fitness value arranged in the first half are regarded as the leader, and the other goblet ascidians are regarded as the follower.
And 8: and (4) updating the position. And updating the position of the leader according to a formula (11), and updating the position of the follower according to a formula (12).
And step 9: and (5) repeating the step (5) to the step (8) until the maximum iteration times or the optimal fitness is reached. The position of the food is the optimal parameter of the GRU model.
Step 10: and (4) carrying out situation prediction on the test set by using the trained SSA-GRU neural network, and verifying the training effect.
In an open SDN architecture, since the upper layer control logic is separated from the underlying infrastructure and the controller has a global view, the SDN controller can be monitored, enabling acquisition of a large amount of real-time and historical data. When the SDN is applied to a large-scale wide area network, the performance of a single SDN controller cannot meet the management requirement of the whole network, and a control plane needs to be expanded and multiple controllers working in cooperation are deployed.
The network flow situation prediction method can comprehensively predict the development trend of the network flow according to the current and historical network flow states, predict the change trend of the network flow in advance and provide decision support for a network administrator to deploy and regulate the network in time. The network flow situation prediction is that on the premise that a network flow situation value is calculated through situation evaluation, the development trend of the future network situation is predicted by combining a real-time situation value and a historical situation value according to a certain prediction method, and then the network flow situation value in a period of time in the future is calculated, so that the change situation of the future network flow state is described quantitatively, a network administrator can better know and understand the network flow state at the next moment, and prospective information of the network flow state is provided for the administrator. Network situational prediction can be abstracted as a set of time series data at fixed time intervals. By establishing a reliable model, an evolution rule is found out from a time sequence of the prediction index, the future development trend of the prediction index is quantitatively estimated, and the index change condition in the near period of time in the future is predicted.
Example two
The present embodiment aims to provide a GRU-based multi-domain SDN network traffic situation prediction system, including: a GRU model configured to:
extracting key elements influencing network flow from the historical flow situation value; then, carrying out weighting quantitative calculation on the flow situation value according to related factors, and evaluating the current flow state;
carrying out situation prediction on the test set by using the trained SSA-GRU neural network;
the GRU model comprises an SDN domain, a database, a situation evaluation server and a situation prediction server, wherein the SDN domain adopts a vertical architecture, the database stores flow situation element values, situation values and prediction value data, and the situation evaluation server and the situation prediction server are used for evaluating the network flow situation in the current time period and predicting the network flow situation in the next time period.
The situation evaluation server obtains a numerical value of the network flow situation in the current time period by carrying out weighted quantitative calculation on the flow situation element values in the database, writes the situation value in the database and transmits the situation value to the situation prediction server, so that the current flow state of the multi-domain SDN network is quantitatively described, and the current network flow situation is determined.
The situation prediction server predicts the change trend of the network traffic by using historical and current multi-domain SDN network traffic situation values, wherein the historical moment situation values are provided by a database, the current moment situation values are provided by a situation evaluation server, and the predicted values are written into the database and transmitted into an inter-domain controller;
the network flow situation prediction adopts a method of combining an SSA algorithm and a GRU algorithm, namely, the SSA algorithm is introduced on the basis of the advantage characteristics of the GRU.
Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using general purpose computer means, or alternatively, they can be implemented using program code that is executable by computing means, such that they are stored in memory means for execution by the computing means, or they are separately fabricated into individual integrated circuit modules, or multiple modules or steps of them are fabricated into a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.
Claims (7)
1. The method for predicting the situation of the flow of the multi-domain SDN network based on the GRU is characterized by comprising the following steps:
collecting historical flow situation values, and constructing a sample set in a sliding window mode;
carrying out data normalization on the sample data, and cutting the sample data into a training set and a test set;
establishing a GRU model, and optimizing GRU model parameters by adopting an SSA algorithm; the GRU model is composed of an SDN domain, a database, a situation evaluation server and a situation prediction server, wherein the SDN domain adopts a vertical architecture, the database stores flow situation element values, situation values and prediction value data, and the situation evaluation server and the situation prediction server are used for evaluating the network flow situation in the current time period and predicting the network flow situation in the next time period;
the multi-domain SDN network adopts a vertical architecture and comprises inter-domain controllers and intra-domain controllers, intra-domain data are collected by the intra-domain controllers, and inter-domain data are collected by the inter-domain controllers;
the situation prediction server predicts the change trend of the network traffic by using historical and current multi-domain SDN network traffic situation values, wherein the historical moment situation values are provided by a database, the current moment situation values are provided by a situation evaluation server, and the predicted values are written into the database and transmitted into an inter-domain controller; the network flow situation prediction adopts a method of combining an SSA algorithm and a GRU algorithm, namely, the SSA algorithm is introduced on the basis of the advantages and the characteristics of the GRU, the SSA algorithm is used for optimizing the weight of the GRU algorithm, and the optimized search direction is controlled;
the GRU model is configured to: extracting key elements influencing network flow from historical flow situation values; then, carrying out weighted quantitative calculation on the flow situation value according to related factors, and evaluating the current flow state; and (4) carrying out situation prediction on the test set by using the trained SSA-GRU neural network.
2. The method of GRU-based multi-domain SDN network traffic situation prediction method of claim 1, wherein extracting traffic situation elements: the controller is connected with each switch, so that a large amount of real-time and historical flow data are obtained, the collected original data are analyzed and processed, key elements influencing the network flow state are extracted, and a uniform data format is formed and stored in a database.
3. The method as claimed in claim 1, wherein the situation assessment server obtains a value of the network traffic situation in the current time period by performing weighted quantitative calculation on the traffic situation element values in the database, writes the current time period situation value into the database and transmits the current time period situation value to the situation prediction server, so as to implement quantitative description on the current traffic state of the multi-domain SDN network, thereby determining the current network traffic situation.
4. The method of claim 1, wherein traffic situation levels of a next period of time are divided into a plurality of classes according to a standard corresponding to a predicted traffic situation value and a traffic situation value;
and aiming at the predicted traffic situation level of the multi-domain SDN network, planning and deciding the multi-domain SDN network in advance.
5. A GRU-based multi-domain SDN network traffic situation prediction system for implementing the GRU-based multi-domain SDN network traffic situation prediction method of any one of claims 1 to 4, comprising: the GRU model comprises an SDN domain, a database, a situation assessment server and a situation prediction server;
the SDN domain adopts a vertical architecture, a database stores flow situation element values, situation values and predictive value data, and a situation evaluation server and a situation predictive server are used for evaluating the network flow situation of the current time period and predicting the network flow situation of the next time period.
6. The GRU-based multi-domain SDN network traffic situation prediction system of claim 5, wherein the situation assessment server obtains a value of the network traffic situation in the current time period by performing weighted quantitative calculation on the traffic situation element values in the database, writes the current time period situation value into the database and transmits the current time period situation value to the situation prediction server, so as to realize quantitative description on the current traffic state of the multi-domain SDN network, thereby determining the current network traffic situation.
7. The GRU-based multi-domain SDN network traffic situation prediction system of claim 5, wherein the situation prediction server predicts a trend of change of the network traffic using historical and current multi-domain SDN network traffic situation values, wherein the historical moment situation values are provided by a database and the current moment situation values are provided by a situation assessment server, and writes the predicted values to the database and into the inter-domain controller;
the network flow situation prediction adopts a method of combining an SSA algorithm and a GRU algorithm, namely, the SSA algorithm is introduced on the basis of the advantage characteristics of the GRU.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010129348.6A CN111371644B (en) | 2020-02-28 | 2020-02-28 | Multi-domain SDN network traffic situation prediction method and system based on GRU |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010129348.6A CN111371644B (en) | 2020-02-28 | 2020-02-28 | Multi-domain SDN network traffic situation prediction method and system based on GRU |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111371644A CN111371644A (en) | 2020-07-03 |
| CN111371644B true CN111371644B (en) | 2023-04-07 |
Family
ID=71208302
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010129348.6A Active CN111371644B (en) | 2020-02-28 | 2020-02-28 | Multi-domain SDN network traffic situation prediction method and system based on GRU |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111371644B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112202619A (en) * | 2020-10-12 | 2021-01-08 | 浪潮云信息技术股份公司 | Intelligent cloud computing network flow adjusting and optimizing system and method |
| CN112235164A (en) * | 2020-11-05 | 2021-01-15 | 全球能源互联网研究院有限公司 | Neural network flow prediction device based on controller |
| CN112118143B (en) * | 2020-11-18 | 2021-02-19 | 迈普通信技术股份有限公司 | Traffic prediction model training method, traffic prediction method, device, equipment and medium |
| CN112989693A (en) * | 2021-03-02 | 2021-06-18 | 上海电机学院 | Wind power prediction method based on SSA-GRU-MSAR |
| CN113036762B (en) * | 2021-05-10 | 2022-05-24 | 中国建筑第五工程局有限公司 | SSA algorithm-based multi-channel power signal mode rapid identification method |
| CN114168967A (en) * | 2021-12-08 | 2022-03-11 | 重庆大唐国际彭水水电开发有限公司 | Industrial control system security situation prediction method and system |
| CN114338378B (en) * | 2022-01-11 | 2023-01-10 | 海南大学 | Configuration verification method for multi-domain software defined network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105491013A (en) * | 2015-11-20 | 2016-04-13 | 电子科技大学 | Multi-domain network security situation perception model and method based on SDN |
| CN109325638A (en) * | 2018-11-09 | 2019-02-12 | 电子科技大学 | A kind of SDN method for predicting based on RBF neural |
| CN109816095A (en) * | 2019-01-14 | 2019-05-28 | 湖南大学 | Based on the network flow prediction method for improving gating cycle neural network |
-
2020
- 2020-02-28 CN CN202010129348.6A patent/CN111371644B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105491013A (en) * | 2015-11-20 | 2016-04-13 | 电子科技大学 | Multi-domain network security situation perception model and method based on SDN |
| CN109325638A (en) * | 2018-11-09 | 2019-02-12 | 电子科技大学 | A kind of SDN method for predicting based on RBF neural |
| CN109816095A (en) * | 2019-01-14 | 2019-05-28 | 湖南大学 | Based on the network flow prediction method for improving gating cycle neural network |
Non-Patent Citations (3)
| Title |
|---|
| SDN网络流量态势评估及预测技术研究;胡孟婷;《中国优秀硕士学位论文全文数据库》;20191215;第I139-44页 * |
| 基于MGU的大规模IP骨干网络实时流量预测;郭芳等;《山东大学学报(工学版)》;20190312(第02期);第88-95页 * |
| 基于改进樽海鞘优化BP神经网络的土壤墒情预测;安小宇等;《中国农机化学报》;20191115(第11期);第124-130页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111371644A (en) | 2020-07-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111371644B (en) | Multi-domain SDN network traffic situation prediction method and system based on GRU | |
| Wang et al. | Machine learning for networking: Workflow, advances and opportunities | |
| CN111526096B (en) | Intelligent identification network state prediction and congestion control system | |
| CN109981691B (en) | SDN controller-oriented real-time DDoS attack detection system and method | |
| Tang et al. | Virtual network function migration based on dynamic resource requirements prediction | |
| Subramanya et al. | Machine learning-driven scaling and placement of virtual network functions at the network edges | |
| CN113114722B (en) | Virtual network function migration method based on edge network | |
| CN113869521A (en) | Method, device, computing equipment and storage medium for constructing prediction model | |
| Wang et al. | xnet: Improving expressiveness and granularity for network modeling with graph neural networks | |
| Babayigit et al. | Deep learning for load balancing of SDN‐based data center networks | |
| CN113887748B (en) | Online federal learning task allocation method and device, and federal learning method and system | |
| Zhou et al. | Multi-task deep learning based dynamic service function chains routing in SDN/NFV-enabled networks | |
| Jiang et al. | ML-based pre-deployment SDN performance prediction with neural network boosting regression | |
| CN104092503A (en) | Artificial neural network spectrum sensing method based on wolf pack optimization | |
| KR20220014744A (en) | Data preprocessing system based on a reinforcement learning and method thereof | |
| Lu et al. | A multi-agent adaptive traffic signal control system using swarm intelligence and neuro-fuzzy reinforcement learning | |
| CN116455820A (en) | Multi-transmission path adjustment system and method based on congestion avoidance | |
| Andrushchak et al. | Intelligent traffic engineering for future intent-based software-defined transport network | |
| CN111327494B (en) | Multi-domain SDN network flow situation assessment method and system | |
| CN117014355A (en) | TSSDN dynamic route decision method based on DDPG deep reinforcement learning algorithm | |
| Afifi et al. | Machine Learning with Computer Networks: Techniques, Datasets and Models | |
| Abdellah et al. | Artificial intelligence driven 5g and beyond networks | |
| St‐Onge et al. | NFVLearn: A multi‐resource, long short‐term memory‐based virtual network function resource usage prediction architecture | |
| Zitouna | Learning-based Orchestrator for Intelligent Software-defined Networking Controllers | |
| Leyva-Pupo et al. | An intelligent scheduling for 5G user plane function placement and chaining reconfiguration |
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 | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230726 Address after: 13F, Hongri Building, 200m north of the intersection of Beijing Road and Shuhe Road, Liuqing Street, Lanshan District, Linyi City, Shandong Province 276000 Patentee after: Government Energy Finance and Taxation (Shandong) Cloud Technology Co.,Ltd. Address before: 264005 191 Binhai Middle Road, Laishan District, Yantai, Shandong Patentee before: SHANDONG INSTITUTE OF BUSINESS AND TECHNOLOGY |