CN113037425B - Multi-target controller placement method based on evolution perception in network - Google Patents

Multi-target controller placement method based on evolution perception in network Download PDF

Info

Publication number
CN113037425B
CN113037425B CN202110284085.0A CN202110284085A CN113037425B CN 113037425 B CN113037425 B CN 113037425B CN 202110284085 A CN202110284085 A CN 202110284085A CN 113037425 B CN113037425 B CN 113037425B
Authority
CN
China
Prior art keywords
population
individual
network
num
controller
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
Application number
CN202110284085.0A
Other languages
Chinese (zh)
Other versions
CN113037425A (en
Inventor
徐展琦
李兴
朱宇豪
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xidian University
Original Assignee
Xidian University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xidian University filed Critical Xidian University
Priority to CN202110284085.0A priority Critical patent/CN113037425B/en
Publication of CN113037425A publication Critical patent/CN113037425A/en
Application granted granted Critical
Publication of CN113037425B publication Critical patent/CN113037425B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04J—MULTIPLEX COMMUNICATION
    • H04J14/00—Optical multiplex systems
    • H04J14/02—Wavelength-division multiplex systems
    • H04J14/0201—Add-and-drop multiplexing
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04J—MULTIPLEX COMMUNICATION
    • H04J14/00—Optical multiplex systems
    • H04J14/02—Wavelength-division multiplex systems
    • H04J14/0201—Add-and-drop multiplexing
    • H04J14/0202—Arrangements therefor

Abstract

The invention discloses a method for placing a multi-target controller based on evolution perception in a network, which mainly solves the problems of overlong propagation delay, large placing cost and large network load difference caused by placing a controller in a control network in the prior art. The implementation scheme is as follows: initializing network topology information and algorithm setting information; preprocessing the initialized information; generating an initialization population, and representing a controller deployment scheme by each individual in the current population; iteratively evolving the current population; checking the legality of the population, and judging all legal individuals and all illegal individuals; correcting illegal individuals to obtain an approximate optimal controller deployment scheme; the placement of the controller is performed in the control network according to one of a set of placement scenarios. The invention can more fully and reasonably utilize network resources, reduce propagation delay and placement cost, improve network performance, and can be used in DWDM wide area optical network.

Description

Multi-target controller placement method based on evolution perception in network
Technical Field
The invention relates to the field of network communication, in particular to a method for placing a multi-target controller, which can be used in a DWDM wide area optical network.
Background
For a long time, the dense wavelength division multiplexing DWDM-based wide area network has the advantages of large transmission capacity, easy transmission of various signals, dynamic reconfiguration of networking and the like, is widely used in actual networks such as core networks, metropolitan area networks, access networks and the like, and advanced new technologies such as data center networks, virtualization technologies, cloud computing, edge computing, internet of things and the like, and can greatly meet related business requirements. With the development of digital services and the popularization of mobile office, the demand of cloud computing services is rapidly increased, information and communication technology ICT service providers construct cloud data centers to provide various cloud computing services, and a cloud data center network needs a wide area network with wider coverage, higher bandwidth, lower time delay and higher quality, which puts higher requirements on the architecture, performance and operation of the wide area network. In addition, due to the dramatic increase in the demand of virtual services, how to connect and manage network devices in a more flexible and reliable manner is becoming a clear demand for virtual networks.
In order to meet the requirements of the cloud era internet service on the service mode and the operation mode of the wide area network, the introduction of a software defined network SDN in the DWDM wide area network based on dense wavelength division multiplexing brings about four great advantages. And firstly, flow balance scheduling is facilitated. The wide area network based on the DWDM technology can utilize a Software Defined Network (SDN) to intelligently and globally control, collect state information of each node in the wide area network in real time, dynamically adjust service routing and change service bandwidth to avoid network congestion, and therefore load balancing of transmission link flow of the wide area network can be achieved. And secondly, the utilization rate of the wide area network link can be improved. Because the wide area network based on the DWDM technology is improved through a Software Defined Network (SDN) and then controls the service flow in a unified centralized and optimized mode, the waste of controller resources or the probability of insufficient controller resources can be reduced. And thirdly, the service quality of the wide area network is effectively improved. The control layer of the SDN can provide a global network view to the application service layer and aggregate a plurality of technologies to provide services, so that the optimal link is selected between a user side and an Internet Data Center (IDC) and a specific server, and higher service quality can be obtained. Fourthly, the performance of the virtual network VN is effectively improved. The SDN is combined with the NFV, so that service requests can be processed in parallel, the peak network demand can be met, resources can be released at any time according to the network demand, and fault management and rapid upgrading of the VN are facilitated.
The software defined network SDN is mainly characterized in that control and forwarding are separated, namely distributed control is changed into control with a plurality of controllers centralized relatively, open standard interfaces are adopted among layers, and necessary forwarding is completed by adopting general hardware. To implement the application of software defined network SDN technology to a wide area network, a certain number of controllers should first be placed on the wide area network, which is usually divided into a plurality of domains, each of which is managed by a separate controller, while the controllers are located to meet specific network requirements, a problem known as controller placement problem SDN-CPPs, which is currently considered as a difficult problem of non-deterministic polynomial NPs.
The SDN-CPPs for solving the problem of placing the controllers are prerequisites for improving the fault tolerance capability and the expandability of the network, the quantity and the positions of the controllers can influence various indexes of the wide area network such as time delay, reliability, elasticity and the like, and the indexes can be abstracted into mutually-constrained optimization targets. In addition, since the network function virtualization NFV is responsible for virtualization of various network elements, and the software defined network SDN is responsible for virtualization of the interconnection between the networks themselves, such as network nodes, the solution of the controller placement problem SDN-CPPs is beneficial to the optimization of the virtual network VN.
Most of the current research only relates to the propagation delay from the controller to the controlled switch, but the propagation delay between the controllers is not considered, so that the delay consumed by propagation of control information between the controllers cannot be optimized, and the performance of the network is further influenced; meanwhile, as the load balance problem of the controllers is rarely researched at present, the load of most of the controllers is too large or too small, so that the storage resources of the controllers in the final placement scheme are insufficient or wasted; furthermore, since no relevant research is concerned about the cost of placement at present, this may result in an excessively high cost of the final placement solution, which is not favorable for practical application of the placement solution.
Disclosure of Invention
The invention aims to provide a multi-target controller placement method based on evolution sensing in a network aiming at the defects of the prior art, so as to improve the performance of a control network, reduce the propagation delay of the control network, promote the reasonable utilization of the storage resources of a controller and optimize the placement cost of the control network.
In order to achieve the purpose, the technical idea of the invention is as follows: the propagation delay of the control network is obtained by weighting and summing the propagation delay from the controller to the controlled switch and the propagation delay between the controllers, so that the delay performance of the control network is more comprehensively optimized; calculating the load difference of the controller through the extreme difference of the number of the switches to which the controller belongs so as to promote the reasonable utilization of the storage resources of the controller; the placement cost of the control network is balanced by reducing the number of controllers to continuously optimize the placement scheme in subsequent operations, which is implemented by the following steps:
(1) initialize Internet OS3E network information and method settings information:
initializing Internet OS3E network information, including setting network node longitudeλiAnd latitude
Figure BDA0002979694910000021
Network topology node total number N and source node A of direct linkiAnd a sink node OiI is 1,2, … …, N;
initializing set information, including setting population size Num, upper limit of controller number maxCNum, evolution perception threshold tau, two differential evolution step length F1And F2Neighborhood range delta, weight vector f and maximum evolution algebra Gmax；
(2) Calculating the length L (A) of the links in the Internet OS3E network topologyi,Oi)；
(3) Generating an initial population:
3a) generating an initial population P1 of the first half scale according to the initialization setting information;
3b) according to the length L (A) of the link in the Internet OS3E network topologyi,Oi) Initializing setting information to generate an initial population P2 of the second half scale;
3c) combining the initial populations P1 and P2 generated by 3a) and 3b) to obtain a complete initial population: p0P1 ═ u P2, in the current population
Figure BDA0002979694910000031
Each representing a controller deployment scenario in which GeneFor the current evolution algebra, GeneIs 0,1, 2, … …, Gmax；
(4) Iterative evolution of the current population:
4a) for the current population
Figure BDA0002979694910000032
Performing non-dominant sorting to obtain the current population
Figure BDA0002979694910000033
Pareto solution set of;
4b) calculating an evolutionary algebra perception coefficient g;
4c) the current population
Figure BDA0002979694910000034
The coding mode is converted from binary coding to Gray code to obtain the current population in the form of Gray code
Figure BDA0002979694910000035
4d) Current population in the form of a corresponding gray code
Figure BDA0002979694910000036
Evolution is carried out:
4d1) judging whether the evolution algebraic perception coefficient g obtained in the step 4b) is less than or equal to an evolution perception threshold tau: if so, perform 4d2), otherwise, perform 4d 3);
4d2) evolving a current population in Gray code form using an evolutionary algorithm based on non-dominated sorting
Figure BDA0002979694910000037
4d3) Evolving current populations in Gray code form using multi-objective decomposition-based algorithms
Figure BDA0002979694910000038
5) For the current population
Figure BDA0002979694910000039
And (3) carrying out validity check, and correcting illegal individuals:
5a) the total number of controllers that need to be deployed for the deployment scenario represented by all individuals, C, is calculated and compared to the number of Internet OS3E network nodes:
if C is larger than half of the total number of the network nodes of Internet OS3E, judging the illegal individual U, and correcting the illegal individual U to obtain a corrected individual U';
if C is less than or equal to half of the total number of the Internet OS3E network nodes, judging the legality of the next individual, and executing 5b) until the legality of all the individuals is judged;
5b) and (3) iteration stop judgment: if the current evolution algebra reaches GeneTo the maximum algebra GmaxIf so, terminating the iteration to obtain the next generation population
Figure BDA0002979694910000041
Executing (6), otherwise, returning to (4);
(6) according to the next generation population
Figure BDA0002979694910000042
And calculating to obtain an approximate optimal controller placement scheme set R, and placing the controller according to each scheme in the placement scheme set.
Compared with the prior art, the invention has the following advantages:
firstly, the initial population P1 of the first half scale is generated according to the initialization setting information, so that the diversity of the initial population is improved; due to the length L (A) of the links in the network topology according to Internet OS3Ei,Oi) Initializing the setting information to generate an initial population P2 of the second half scale, thereby enhancing the convergence of the initial population; in addition, the two parts of initial populations are combined to obtain a complete initial population, so that an initial population with better diversity and convergence can be obtained, and convenience is provided for the next step of evolution operation.
Secondly, a more appropriate evolution mode is selected according to the evolution algebra perception coefficient g, and when the evolution algebra perception coefficient g is smaller, the current population in the form of the Gray code is evolved by using an evolution algorithm based on non-dominated sorting
Figure BDA0002979694910000043
The distribution of the population is improved; when the evolution algebra perception coefficient g is larger, the current population in the form of Gray code is evolved by using a multi-objective decomposition algorithm
Figure BDA0002979694910000044
The convergence of the population is improved, so that the population can be promoted to evolve towards a better direction, the performance of a higher control network can be conveniently obtained, and the lower control is realizedThe controller placement scheme set is used for controlling network propagation delay, more reasonably configuring the storage resources of the controller and reducing the placement cost of the controller.
Third, the present invention is in the current population
Figure BDA0002979694910000045
Before evolution, individual coding modes are converted into Gray codes, so that the global search capability of the current population is enhanced, and the current population is prevented from being searched
Figure BDA0002979694910000046
Trapping into a local optimal solution; in the current population
Figure BDA0002979694910000047
After evolution, the coding mode of the individual is converted into binary coding, thereby providing convenience for the selection of the next generation of individual.
Drawings
FIG. 1 is a general flow chart of an implementation of the present invention;
FIG. 2 is a flow chart of an initial population generation sub-process in the present invention;
FIG. 3 is a flow chart of a mixed-evolutionary algorithm of the present invention;
FIG. 4 is a comparison graph of the optimized target values of the set of near-optimal controller placement solutions R obtained under the Internet OS3E using the present invention and the comparison method.
Detailed Description
Embodiments and effects of the present invention are described in further detail below with reference to the accompanying drawings.
Referring to fig. 1, the present example is implemented based on various improved mechanisms, and the specific steps are as follows:
step 1: the Internet OS3E network information and method setting information are initialized.
1.1) initializing Internet OS3E network information, which includes setting network node longitude λiAnd latitude
Figure BDA0002979694910000051
Network topology node total number N, source node A of direct linkiAnd a sink node OiI is a number of 1,2,……，N；
1.2) initializing the setting information, including setting the population size Num, the upper limit of the controller number maxCNum, the evolution perception threshold tau, and two differential evolution steps F1And F2Neighborhood range delta, weight vector f and maximum evolution algebra Gmax。
The example is set according to simulation experiments, but not limited to, Num 100, evolution perception threshold τ 0.4, maximum evolution algebra Gmax100, 34 of total number of network topology nodes N, and differential evolution step length F10.3 and F20.6, 10 neighborhood range δ, 30 weight vector Γ.
Step 2: calculating the length L (A) of the links in the Internet OS3E network topologyi,Oi) And finds the shortest path between any two nodes in the Internet OS3E network topology.
The Internet OS3E network, an american backbone, is widely used for simulation studies of controller placement problems where link lengths need to be calculated.
2.1) calculating the length of the link in the Internet OS3E network topology according to the initialization parameters by using the following hemiversine formula:
Figure BDA0002979694910000052
in the formula RearthRepresenting the radius of the earth, by a value Rearth＝6400km，
Figure BDA0002979694910000053
Represents the ith source node AiThe longitude of (a) is determined,
Figure BDA0002979694910000054
represents the ith source node AiThe latitude of (a) is determined,
Figure BDA0002979694910000055
represents the ith sink node OiThe longitude of (a) is determined,
Figure BDA0002979694910000056
represents the ith sink node OiLatitude, L (A)i,Oi) Represents the ith source node AiAnd the ith sink node OiThe length of the direct link between the two links, i is 1,2, … …, N;
2.2) find the shortest circuit between any two nodes in the Internet OS3E network topology:
the existing method for determining the shortest path between any two nodes in the Internet OS3E network topology includes a Floyd algorithm and a Dijkstra algorithm, and the present embodiment selects, but is not limited to, the Floyd algorithm, which is implemented as follows:
will be input into the Internet OS3E network topology network ith node longitude lambdaiLatitude and longitude
Figure BDA0002979694910000057
Number N of Internet OS3E network topology nodes, ith source node AiAnd the ith sink node OiLength of direct link between L (A)i,Oi) These parameters are input to the algorithm Floyd algorithm;
calculating and outputting a shortest circuit set S between any two nodes i and d in the Internet OS3E network topology through MATLAB softwareidI, d ≠ 1,2, … …, N, and i ≠ d.
And step 3: and generating an initial population.
Referring to fig. 2, the specific implementation of this step is as follows:
3.1) generating the initial population P1 of the first half scale according to the initialization setting information:
3.1.1) calculating the group number setIndex (q) of the qth individual in the first half-scale initial population P1:
Figure BDA0002979694910000061
wherein q is 1,2, …, Num' represents half of population scale, and takes value
Figure BDA0002979694910000062
3.1.2) setting the locus judgment threshold value alpha (q) of the qth individual according to the group number setIndex (q):
Figure BDA0002979694910000063
3.1.3) determining the ith locus P (q, i) of the individual q on the basis of the locus discrimination threshold α (q):
Figure BDA0002979694910000064
wherein, er represents a random number ranging between (0,1), all gene positions are sequentially determined, all gene positions P (q, i) of an individual q are determined, i is 1,2, …, and N is the individual q;
3.1.4) repeating steps 3.1.1) -3.1.3) until q equals Num', generating the first half-size initial population P1.
3.2) generating the second half-scale initial population P2:
3.2.1) randomly generating a random number Ccenter in the range of (0, maxNum)qNumber of clustering centers Ccenter for the qth individualq，q＝1,2,…,Num'；
3.2.2) input Internet OS3E network topology network node longitude λ using k-means clustering algorithmiLatitude and longitude
Figure BDA0002979694910000065
Internet OS3E network topology node number N, and clustering center number CcenterqSource node AiAnd a sink node OiLength of direct link between L (A)i,Oi) And outputting a controller deployment scheme, namely finishing the initialization of the qth individual, wherein the value of i is 1,2, … …, N,
3.2.3) repeating steps 3.2.1) -3.2.2) until q equals Num', the second half-size initial population P2 is generated.
3.3) combining the initial populations P1 and P2 generated in 3.1) and 3.2), namely taking the union of the two initial populations to obtain the complete populationInitial population of (a): p0＝P1∪P2。
And 4, step 4: and (5) iteratively evolving the current population.
Referring to fig. 3, the specific implementation of this step is as follows:
4.1) to the current population
Figure BDA0002979694910000071
Performing non-dominant sorting to obtain the current population
Figure BDA0002979694910000072
Pareto solution set of; in the current population
Figure BDA0002979694910000073
Each representing a controller deployment scenario in which GeneFor the current evolution algebra, GeneIs 0,1, 2, … …, Gmax；
4.2) calculating an evolutionary algebraic perceptual coefficient g by the following formula:
Figure BDA0002979694910000074
wherein G iseneRepresenting the current evolution algebra, GmaxRepresenting the maximum evolutionary algebra.
4.3) matching the current population
Figure BDA0002979694910000075
The coding mode is converted from binary coding to Gray code to obtain the current population in the form of Gray code
Figure BDA0002979694910000076
4.4) Current population in the form of Pair Gray codes
Figure BDA0002979694910000077
Evolution is carried out:
4.4.1) judging whether the evolution algebraic perception coefficient g obtained in the step 4.2) is less than or equal to the evolution perception threshold tau:
if so, then 4.4.2 is executed),
otherwise, execute 4.4.3);
4.4.2) evolution of the current population in Gray code form using an evolutionary algorithm based on non-dominated sorting
Figure BDA0002979694910000078
4.4.2a) sequentially from the current population in Gray code form
Figure BDA0002979694910000079
Randomly selects two parent individuals y1 in the nt groupntAnd y2ntAnd performing single-point crossing and single-point mutation on the two parent individuals to obtain a child individual y1rt' and y2nt'，nt＝1，2，……，Num；
4.4.2b) repeat step 4.4.2a) until nt ═ Num, generating the offspring population PG of size Num gray code formoff＝{y1nt',nt＝1,2,…,Num}；
4.4.2c) Subsubstitution PG in Gray code formoffThe encoding mode of the method is converted from Gray code encoding into binary encoding to obtain a filial generation population PoffAnd comparing the current population
Figure BDA0002979694910000081
And a progeny population PoffThe whole population is obtained by merging,
Figure BDA0002979694910000082
get the current population immediately
Figure BDA0002979694910000083
And a progeny population PoffA union of (1);
4.4.2d) Overall population
Figure BDA0002979694910000084
Performing non-dominant sorting to obtain the next generation population
Figure BDA0002979694910000085
4.4.3) Using Multi-target decomposition Algorithm based on the present population in the form of Gray codes
Figure BDA0002979694910000086
Evolution is carried out:
4.4.3a) from the Current population
Figure BDA0002979694910000087
R1 of the two parent individuals of the nt groupntAnd r2ntAnd from the current population
Figure BDA0002979694910000088
Randomly selecting an individual p in the Pareto solution set, and generating the nt group of evolved individuals r by the following formulant'：
rnt'＝rnt+F1·(r1nt-r2nt)+F2·(p-r1nt)，
Wherein r isntRepresents the nt group of individuals to be evolved, nt is 1,2, … …, Num;
4.4.3b) Subjects rntThe coding mode of the' is converted from Gray code coding to binary coding to obtain evolved individual offnt,nt＝1,2,…,Num；
4.4.3c) repeat steps 4.4.3a) -4.4.3 b) until nt ═ Num, yielding the next generation population
Figure BDA00029796949100000811
And 5: for the next generation population
Figure BDA0002979694910000089
And carrying out validity check to obtain the corrected individual U'.
5.1) calculating the total number C of the controllers needing to be deployed in the deployment scheme represented by all the individuals:
5.1.1) from the Next Generation population
Figure BDA00029796949100000810
The nth individual is taken out, and the total controller number C required by the deployment scheme represented by the nth individual is calculatednt：
Cnt＝sum(nt)，nt＝1,2,…,Num；
5.1.2) repeating the step 5.1.1) until nt ═ Num, obtaining the total number of controllers which need to be deployed for the deployment scheme represented by all individuals: c ═ Cnt，nt＝1,2,…,Num}；
5.2) verifying the legality of all individuals in the current population:
the total number C of the controllers which need to be deployed in the deployment scheme represented by all the individuals is compared with the number of Internet OS3E network nodes:
if C is larger than half of the total number of the network nodes of Internet OS3E, judging the illegal individual U, correcting the illegal individual U, and executing 5.3);
if C is less than or equal to half of the total number of the Internet OS3E network nodes, judging the legality of the next individual, and executing 5.4) until the legality of all the individuals is judged;
5.3) correcting illegal individuals:
5.3.1) calculating the ith locus Dis (i) of the perturbation vector:
Figure BDA0002979694910000091
5.3.2) according to different gene positions Dis (i) and gene positions U (i) of the illegal individuals, the gene positions of the illegal individuals are corrected as follows:
Figure BDA0002979694910000092
wherein U ' (i) denotes the ith gene position of the legal individual after correction, U (i) denotes the ith gene position of the illegal individual, U (i) + Dis (i) denotes the sum of the U of the illegal individual and the ith gene position of the perturbation vector Dis, when U (i) + Dis (i) is 1, U ' (i) is 1, that is, the controller is to be placed in the network node corresponding to the ith gene position, otherwise, U ' (i) is 0, that is, the controller is not to be placed in the network node corresponding to the ith gene position;
5.3.3) obtaining the corrected legal individual U ' ═ { U ' (i), i ═ 1,2, …, N } from all the gene positions U ' (i) of the corrected legal individual.
5.4) judging iteration stop: if the current evolution algebra reaches GeneTo the maximum algebra GmaxIf so, terminating the iteration to obtain the next generation population
Figure BDA0002979694910000093
Step 6 is executed, otherwise, the step 4 is returned;
step 6: according to the next generation population
Figure BDA0002979694910000094
And calculating to obtain an approximate optimal controller placement scheme set R.
6.1) for the next generation population
Figure BDA0002979694910000095
Performing non-dominant sorting to obtain the next generation population
Figure BDA0002979694910000096
A corresponding Pareto solution set, namely an approximate optimal controller placement scheme set R;
6.2) selecting kth individual x from the approximate optimal controller placement solution set RktRespectively connecting kth individual xktThe controller placement scheme represented is applied to the Internet OS3E network, kt ═ 1,2, … …, | R |;
6.3) repeat step 6.2) until kt ═ R |, completing the application of all controller placement schemes.
The effects of the present invention can be further illustrated by the following simulations:
firstly, simulation conditions:
condition 1: the population scale Num of the multi-target differential evolution algorithm MOEA/D based on decomposition is set as 100, and an evolution algebra Gmax100, cross probabilityPc0.9, probability of mutation Pm0.9, difference coefficient F10.3 and F2＝0.3；
Condition 2: population size Num of multi-objective evolutionary algorithm NSGA-II (non-dominated sorting) based on elite selection strategy is 100, and evolution algebra Gmax100, cross probability Pc0.9, probability of mutation Pm＝0.9。
Simulation software: MATLAB software was used.
Secondly, simulation content:
simulation 1, simulating the placement positions of the multi-target controller by using the method, solving the obtained approximate optimal controller placement scheme set R, and calculating all optimal placement scheme optimization target values F (x) through the following steps:
firstly, computing kth individual x in an approximate optimal controller placement scheme set RktThe objective function value of (1):
F(xkt)＝(fM(xkt),fD(xkt),fL(xkt))T，
wherein f isM(xkt)＝sum(xkt) Representing kth individual xktThe cost of placement of (1), (2), (… …), (R) |;
fD(xkt)＝max(xkt)-min(xkt) Representing kth individual xktThe load difference of (2);
Figure BDA0002979694910000101
representing kth individual xktMaximum propagation delay, beta1And beta2V represents the average propagation velocity of light in the network for two propagation delay weight coefficients, in this example taken as beta1＝β2＝0.5，v＝2×108m/s, i, d ≠ 1,2, … …, N, and i ≠ d;
repeating the calculation until kt is equal to | R |, and obtaining the objective function value F of all placement schemes of the simulation approximate optimal controller placement scheme set R1(x)＝{F(xkt) Is F of the formula1(x) Is aAnd 108 rows and 3 columns of matrix.
Simulating 2, simulating the placement positions of the multi-target controllers by using the existing multi-target differential evolution algorithm MOEA/D based on decomposition under the condition 1, solving the obtained approximate optimal controller placement scheme set R, and calculating all optimal placement scheme optimization target values F of the approximate optimal controller placement scheme set R2(x) The calculation procedure is the same as in simulation 1. Calculated F2(x) Is a matrix of 95 rows and 3 columns.
And 3, simulating the placement positions of the multi-target controllers by using the existing multi-target evolutionary algorithm NSGA-II based on non-dominated sorting and elite selection strategies under the condition 2, solving the obtained approximately optimal controller placement scheme set R, and calculating all optimal target values F of the placement schemes3(x) The calculation procedure is the same as in simulation 1. Calculated F3(x) Is a matrix of 101 rows and 3 columns
Optimizing target values F of three all placement schemes obtained by simulation 1, simulation 2 and simulation 31(x)、F2(x) And F3(x) Inputting the data into MATLAB software, and drawing in the MATLAB software to obtain an optimized target value comparison graph of the approximate optimal controller placement scheme set R obtained by the two methods under the Internet OS3E, as shown in FIG. 4, wherein the abscissa is placement cost, the ordinate is load difference, and the ordinate is maximum propagation delay.
As can be seen from fig. 4, compared with the prior art, the method of the present invention can obtain a shorter maximum propagation delay under the same placement cost and load difference, and when the placement cost is determined, the method of the present invention obtains a placement scheme set with a lower maximum propagation delay and load difference. This shows that the approximate optimal controller placement scheme set R obtained according to the present invention places the controller into the network, which can improve the performance of the control network, reduce the propagation delay of the control network, promote the reasonable utilization of the controller storage resources, and optimize the placement cost of the control network.

Claims (6)

1. A multi-target controller placement method based on evolution perception in a network is characterized by comprising the following steps:
(1) initialize Internet OS3E network information and method settings information:
initializing Internet OS3E network information, including setting network node longitude λiAnd latitude
Figure FDA0003380849140000011
Network topology node total number N and source node A of direct linkiAnd a sink node OiI is 1,2, … …, N;
initializing set information, including setting population size Num, upper limit of controller number maxCNum, evolution perception threshold tau, two differential evolution step length F1And F2Neighborhood range delta, weight vector f and maximum evolution algebra Gmax；
(2) Calculating the length L (A) of the links in the Internet OS3E network topologyi,Oi)；
(3) Generating an initial population:
3a) generating an initial population P1 of the first half scale according to the initialization setting information; the method is realized as follows:
3a1) the group number setindex (q) of the qth individual in the first half-scale initial population P1 was calculated:
Figure FDA0003380849140000012
wherein q is 1,2, …, Num' represents half of population scale, and takes value
Figure FDA0003380849140000013
3a2) The locus judgment threshold α (q) of the qth individual is set according to the group number setIndex (q):
Figure FDA0003380849140000014
3a3) determining the ith gene position P (q, i) of the individual q according to the gene position judgment threshold value alpha (q):
Figure FDA0003380849140000015
wherein, er represents a random number ranging between (0,1), all gene positions are sequentially determined, all gene positions P (q, i) of an individual q are determined, i is 1,2, …, and N is the individual q;
3a4) repeating steps 3a1) -3 a3) until q is Num', generating initial population P1 of the former half scale;
3b) according to the length L (A) of the link in the Internet OS3E network topologyi,Oi) Initializing setting information to generate an initial population P2 of the second half scale; the method is realized as follows:
3b1) randomly generating a random number Ccenter in the range of (0, maxNum)qNumber of clustering centers Ccenter for the qth individualq，q＝1,2,…,Num'；
3b2) Inputting Internet OS3E network topology network node longitude lambda using k-means clustering algorithmiLatitude and longitude
Figure FDA0003380849140000029
Internet OS3E network topology node total number N, and cluster center number CcenterqSource node AiAnd a sink node OiLength of direct link between L (A)i,Oi) Outputting a controller deployment scheme, namely finishing the initialization of the qth individual, wherein the value of i is 1,2, … …, N;
3b3) repeating steps 3b1) -3 b2) until q is Num', generating initial population P2 of the latter half scale;
3c) combining the initial populations P1 and P2 generated by 3a) and 3b) to obtain a complete initial population: p0P1 ═ u P2, in the current population
Figure FDA0003380849140000021
Each representing a controller deployment scenario in which GeneFor the current evolution algebra, GeneIs 0,1, 2, … …, Gmax；
(4) Iterative evolution of the current population:
4a) for the current population
Figure FDA0003380849140000022
Performing non-dominant sorting to obtain the current population
Figure FDA0003380849140000023
Pareto solution set of;
4b) calculating an evolutionary algebra perception coefficient g;
4c) the current population
Figure FDA0003380849140000024
The coding mode is converted from binary coding to Gray code to obtain the current population in the form of Gray code
Figure FDA0003380849140000025
4d) Current population in the form of a corresponding gray code
Figure FDA0003380849140000026
Evolution is carried out:
4d1) judging whether the evolution algebraic perception coefficient g obtained in the step 4b) is less than or equal to an evolution perception threshold tau: if so, perform 4d2), otherwise, perform 4d 3);
4d2) evolving a current population in Gray code form using an evolutionary algorithm based on non-dominated sorting
Figure FDA0003380849140000027
The method is realized as follows:
4d2a) in turn from the current population in the form of gray codes
Figure FDA0003380849140000028
Randomly selects two parent individuals y1 in the nt groupntAnd y2ntAnd performing single-point crossing and single-point mutation on the two parent individuals to obtain a child individual y1rt' and y2nt'，nt＝1，2，……，Num；
4d2b) repeats step 4d2a) until nt ═ Num, generating the child population PG of the form of Num gray codes of sizeoff＝{y1nt',nt＝1,2,…,Num}；
4d2c) grouping of offspring PG in Gray code formoffThe encoding mode of the method is converted from Gray code encoding into binary encoding to obtain a filial generation population PoffAnd comparing the current population
Figure FDA0003380849140000031
And a progeny population PoffThe whole population is obtained by merging,
Figure FDA0003380849140000032
4d2d) overall population
Figure FDA0003380849140000033
Performing non-dominant sorting to obtain the next generation population
Figure FDA0003380849140000034
4d3) Evolving current populations in Gray code form using multi-objective decomposition-based algorithms
Figure FDA0003380849140000035
The method is realized as follows:
4d3a) from the current population
Figure FDA0003380849140000036
R1 of the two parent individuals of the nt groupntAnd r2ntAnd from the current population
Figure FDA0003380849140000037
Randomly selecting an individual p in the Pareto solution set, and generating the nt group of evolved individuals r by the following formulant'：
rnt'＝rnt+F1·(r1nt-r2nt)+F2·(p-r1nt)，
Wherein r isntRepresents the nt group of individuals to be evolved, nt is 1,2, … …, Num;
4d3b) from the subject rntThe coding mode of the' is converted from Gray code coding to binary coding to obtain evolved individual offnt,nt＝1,2,…,Num；
4d3c) repeating steps 4d3a) -4 d3b) until nt equals Num, and obtaining the next generation population
Figure FDA0003380849140000038
5) For the current population
Figure FDA0003380849140000039
And (3) carrying out validity check, and correcting illegal individuals:
5a) the total number of controllers that need to be deployed for the deployment scenario represented by all individuals, C, is calculated and compared to the number of Internet OS3E network nodes:
if C is larger than half of the total number of the network nodes of Internet OS3E, judging the illegal individual U, and correcting the illegal individual U to obtain a corrected individual U';
if C is less than or equal to half of the total number of the Internet OS3E network nodes, judging the legality of the next individual, and executing 5b) until the legality of all the individuals is judged;
5b) and (3) iteration stop judgment: if the current evolution algebra GeneUp to the maximum algebra GmaxIf so, terminating the iteration to obtain the next generation population
Figure FDA00033808491400000310
Executing (6), otherwise, returning to (4);
(6) according to the next generation population
Figure FDA0003380849140000041
Calculating to obtain an approximate optimal controller placement scheme set R according to placementAnd placing each scheme in the scheme set, and placing the controller.
2. The method of claim 1, wherein the length of the link in the Internet OS3E network topology is calculated in (2) by the following hemiversine formula:
Figure FDA0003380849140000042
in the formula RearthRepresenting the radius of the earth, by a value Rearth＝6400km，
Figure FDA0003380849140000043
Represents a source node AiThe longitude of (a) is determined,
Figure FDA0003380849140000044
represents a source node AiThe latitude of (a) is determined,
Figure FDA0003380849140000045
represents a sink node OiThe longitude of (a) is determined,
Figure FDA0003380849140000046
represents a sink node OiLatitude, L (A)i,Oi) Represents a source node AiAnd a sink node OiThe length of the direct link between the two links, i, is 1,2, … …, N.
3. The method according to claim 1, wherein the algebraic perceptual coefficient g in 4b) is calculated by the following formula:
Figure FDA0003380849140000047
wherein G iseneRepresenting the current evolution algebra, GmaxRepresenting the maximum evolutionary algebra.
4. The method according to claim 1, wherein the total number of controllers C required to be deployed for the deployment scenario represented by all the individuals is calculated in 5a) as follows:
5a1) from the next generation population
Figure FDA0003380849140000048
The nth individual is taken out, and the total controller number C needed by the deployment scheme represented by the nth individual is calculatednt：
Cnt＝sum(nt)，nt＝1,2,…,Num；
5a2) Repeating the step 5a1) until nt equals Num, and obtaining the total number of the controllers which need to be deployed by the deployment scheme represented by all the individuals: c ═ Cnt,nt＝1,2,…,Num}。
5. The method according to claim 1, wherein the illegal individual U is corrected in 5a) to obtain a corrected legal individual U', and the following is achieved:
5a3) by calculating the ith locus Dis (i) of the perturbation vector:
Figure FDA0003380849140000051
5a4) according to different gene positions Dis (i) and gene positions U (i) of illegal individuals, the gene positions of the illegal individuals are corrected as follows:
Figure FDA0003380849140000052
wherein U ' (i) denotes the ith gene position of the legal individual after correction, U (i) denotes the ith gene position of the illegal individual, U (i) + Dis (i) denotes the sum of the U of the illegal individual and the ith gene position of the perturbation vector Dis, when U (i) + Dis (i) is 1, U ' (i) is 1, that is, the controller is to be placed in the network node corresponding to the ith gene position, otherwise, U ' (i) is 0, that is, the controller is not to be placed in the network node corresponding to the ith gene position;
5a5) from all the gene positions U ' (i) of the corrected legal individual, U ' ═ { U ' (i), i ═ 1,2, …, N } of the corrected legal individual is obtained.
6. The method of claim 1, wherein (6) is based on a next generation population
Figure FDA0003380849140000053
Calculating to obtain an approximate optimal controller placement scheme set R, which is used for the next generation of population
Figure FDA0003380849140000054
Performing non-dominant sorting to obtain the next generation population
Figure FDA0003380849140000055
And (3) corresponding Pareto solution sets, namely the Pareto solution sets are approximate optimal controller placement scheme sets R.
CN202110284085.0A 2021-03-17 2021-03-17 Multi-target controller placement method based on evolution perception in network Active CN113037425B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110284085.0A CN113037425B (en) 2021-03-17 2021-03-17 Multi-target controller placement method based on evolution perception in network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110284085.0A CN113037425B (en) 2021-03-17 2021-03-17 Multi-target controller placement method based on evolution perception in network

Publications (2)

Publication Number Publication Date
CN113037425A CN113037425A (en) 2021-06-25
CN113037425B true CN113037425B (en) 2022-03-22

Family

ID=76470842

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110284085.0A Active CN113037425B (en) 2021-03-17 2021-03-17 Multi-target controller placement method based on evolution perception in network

Country Status (1)

Country Link
CN (1) CN113037425B (en)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108400890A (en) * 2018-01-30 2018-08-14 天津城建大学 A kind of software defined network multi-controller dispositions method

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8560647B2 (en) * 2011-07-19 2013-10-15 Telefonaktiebolaget L M Ericsson (Publ) Controller placement for split architecture networks
US8804490B2 (en) * 2011-07-29 2014-08-12 Telefonaktiebolaget L M Ericsson (Publ) Controller placement for fast failover in the split architecture
CN107743147A (en) * 2017-11-06 2018-02-27 柏域信息科技（上海）有限公司 Dynamic cloud storage SDN controllers collocation method and device based on global optimization

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108400890A (en) * 2018-01-30 2018-08-14 天津城建大学 A kind of software defined network multi-controller dispositions method

Also Published As

Publication number Publication date
CN113037425A (en) 2021-06-25

Similar Documents

Publication Publication Date Title
CN111538570B (en) Energy-saving and QoS guarantee-oriented VNF deployment method and device
EP2483850A2 (en) Apparatus and method for determining optimum paths in a multi-layer network using a routing engine
CN112272381B (en) Satellite network task deployment method and system
CN111343651B (en) Service chain deployment method and system for serving crowd-sourcing computing environment
CN109347657B (en) Method for constructing virtual data domain of scientific and technological service under SDN mode
CN113472671B (en) Method, device and computer readable storage medium for determining multicast route
CN116107692A (en) Edge container multi-target ant colony optimization scheduling method for dynamic fusion genetic algorithm
Montana et al. Adaptive reconfiguration of data networks using genetic algorithms
CN113037425B (en) Multi-target controller placement method based on evolution perception in network
Oliveira et al. Joint mapping and routing of virtual network functions for improved disaster recovery support
Rodis et al. Intelligent network service embedding using genetic algorithms
CN113285832B (en) NSGA-II-based power multi-mode network resource optimization allocation method
CN113810916B (en) Multi-server mixed deployment architecture and method in 5G/6G edge computing scene
WO2016083845A1 (en) Determining bandwidth requirements for network services
CN115633083A (en) Power communication network service arrangement method, device and storage medium
CN114785692A (en) Virtual power plant aggregation regulation and control communication network flow balancing method and device
Robbi et al. A modified genetic algorithm for resource allocation in cognitive radio networks
CN110933728B (en) Mapping method, device and storage medium of virtualized wireless sensor network
CN114884818A (en) Three-layer soft slicing system and method based on time slot fine granularity in deterministic network
Boyang et al. Virtual network embedding based on hybrid adaptive genetic algorithm
CN112990608A (en) 5G optimization construction method suitable for full coverage of electric power facilities
Pan et al. A genetic algorithm for the overlay multicast routing problem
Noormohammadpour et al. Efficient inter-datacenter bulk transfers with mixed completion time objectives
CN113452552B (en) Information entropy perception-based super-multi-target controller placement method
CN108737192B (en) Network service deployment method based on service reliability

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