CN102882727B - Monitoring area partition method for hierarchical monitoring network - Google Patents

Abstract

The invention discloses a monitoring area partition method for a hierarchical monitoring network, and belongs to the technical field of computer network monitoring. The method includes: taking minimum network monitoring flow as a target, taking domain management central loading balance as a constraint condition to establish a mathematical model; when the mathematical model is optimized and solved, firstly utilizing the genetic algorithm to determine positions of each domain management center; then distributing other nodes in the network to domain management centers already selected according to delay optimal strategy to partition the whole network into multiple monitoring areas. In the genetic algorithm, binary coding is adopted to determine the positions of each domain management center, a fitness function is designed according to the constraint condition and a target function, and by the crossing and mutation method of adaptive probability, the algorithm can guarantee population diversity at initial stage and can quicken convergence speed of the algorithm at later stage. The monitoring area partition method for the hierarchical monitoring network can reduce flow generated by monitoring, reduce monitoring time and maintain loading balance among areas.

Description

A kind of monitored area division methods of stratification monitoring network

Technical field

The present invention relates to a kind of monitored area division methods being applicable to catenet, particularly relate to a kind of monitored area division methods of stratification monitoring network, belong to computer network supervise technical field.

Background technology

Along with the sharp increase of Internet scale, the topology of network becomes and becomes increasingly complex with irregular, ISP and large enterprise need to dispose an overall network monitoring system in network internal, obtain the performance data of network in real time, the performance of monitoring network and safe condition.Network monitoring system due to the modern times pays attention to the management of seeervice level, application layer, and observation process needs the data acquiring frequency of larger data volume and Geng Gao.Therefore, how can Obtaining Accurate Monitoring Data, the impact of observation process on real network transmission data can be reduced again, namely how to form an effective network monitoring system, to ensure disparate networks application sharing limited network resources, become a root problem of network research.

Existing network monitor architecture mainly comprises centralized resources monitoring model and hierarchical structure two kinds.Wherein, centralized monitoring model easily makes Centroid become system bottleneck, can not be applicable to large scale network well, and Hierarchical resource monitoring model has good extensibility and flexibility, can meet the requirement of large scale network monitoring preferably.But, also there is problems such as monitoring the flow produced is comparatively large, the monitoring response time is long, the load imbalance of regional administrative center in the network monitor system of existing hierarchical structure.

Summary of the invention

Technical problem to be solved by this invention is that the monitoring generation flow overcoming the existence of existing stratification monitoring system is large, the monitoring response time is long, the unbalanced problem of area load, provides a kind of monitored area division methods of stratification monitoring network.

The present invention specifically solves the problems of the technologies described above by the following technical solutions:

A kind of monitored area division methods of stratification monitoring network, described stratification monitoring network is divided into multiple monitored area, each monitored area comprises a Ge Yu administrative center, carry out information interaction with network monitoring center for other nodes of monitoring in this monitored area, the splitting scheme of described monitored area is obtained by the following Mathematical Modeling of Optimization Solution:

$\mathrm{Minf}(x_{\mathrm{ij}},y_i)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}T{f}_{\mathrm{ij}}{x}_{\mathrm{ij}}+\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{y}_i(\mathrm{rtT}{f}_j+2\mathrm{MaTf})$

s.t.

$\underset{j=1}{\overset{\left|V\right|}{\mathrm{\Σ}}}{x}_{\mathrm{ij}}=1,(\∀v_i\∈V_a)$

$\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_{\mathrm{ij}}\≤y_j,j=\mathrm{1,2}...,n$

$\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{y}_i\≤P$

$\underset{e_{\mathrm{ij}}\∈E}{\mathrm{\Σ}}{b}_e^{\mathrm{ij}}D\left(e_{\mathrm{ij}}\right)x_{\mathrm{ij}}\≤\mathrm{\δ},i,j=\mathrm{1,2}...,n$

λ<BThreshold

Wherein:

V={v ₀, v ₁..., v _nrepresent the set of all nodes in network, v ₀represent network monitoring center, n represents monitored network node number; Set V _a( 1<=a<n) the set of representative domain administrative center.E=(e _ijrepresent the set of all links in network, wherein, e _ij=(v _i, v _j) represent node v _iand v _jbetween link;

Tf _ijrepresentative domain administrative center v _jmonitor its management domain interior joint v _itime produce flow;

RtTf _jrepresentative domain administrative center v _jby the flow produced when sending it back Surveillance center after the Information procession of collection;

MaTf represents the monitoring flow that any one node of monitoring is once produced;

P represents the maximum number selecting territory administrative center;

D (e _ij) represent link e _ijdelay;

δ is delay-tolerant coefficient;

λ is the load balancing factor of territory administrative center, is defined as:

$\mathrm{\&lambda;}=\sqrt{\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{|{\mathrm{Load}}_i-\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{Load}}_i}{p}|}^2}{p-1}}$

Wherein, Load _irepresent the load of the i-th Ge Yu administrative center, the actual number of p representative domain administrative center;

BThreshold represents the threshold value of the load balancing factor.

Preferably, to described Mathematical Modeling be optimized solve time, first utilize genetic algorithm to determine the position of each territory administrative center; Then, other peer distribution in network are given according to delay optimal policy the territory administrative center chosen, whole network is divided into multiple monitored area.

Further, when utilizing genetic algorithm to determine the position of each territory administrative center, its mapping relations are as follows:

Each chromosome represents a kind of area division scheme, a solution of corresponding described Mathematical Modeling; Be that each node in network composes one No. ID, as chromosomal gene order number with continuous print integer; Chromosomal length is monitored network node number n; Chromosome is encoded in a binary fashion, and the value of gene is 1, represents that node corresponding to this gene sequence number is chosen as territory administrative center;

In genetic algorithm, evaluate each chromosomal fitness function as follows:

$\mathrm{fit}\left(\mathrm{Ci}\right)=\frac{\mathrm{CTf}\left(\mathrm{Ci}\right)+\mathrm{NTf}\left(\mathrm{Ci}\right)}{\mathrm{MaxTf}}+\frac{\mathrm{\&lambda;}\left(\mathrm{Ci}\right)}{\mathrm{BThreshold}}+\mathrm{Overdelay}\left(\mathrm{Ci}\right)$

Wherein, Ci represents a chromosome, and corresponding area division scheme is ({ x _ij, { y _j), CTf(Ci) represent under the area division scheme that chromosome Ci is corresponding, territory administrative center and network monitoring center mutual time the total flow that produces; Under NTf (Ci) represents the zone scheme that Ci is corresponding, the total flow produced when each node is mutual in territory administrative center and its management domain; MaxTf represents when not zoning, the total flow that monitoring produces; λ (Ci) is under the zone scheme that Ci is corresponding, the load balancing factor of territory administrative center; Overdelay (Ci) is the deferred constraint factor, and its computing formula can be expressed as:

$\mathrm{Overdelay}\left(\mathrm{Ci}\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i\max \{\underset{e_{\mathrm{ij}}\&Element;E}{\mathrm{\&Sigma;}}{b}_e^{\mathrm{ij}}D\left(e_{\mathrm{ij}}\right)x_{\mathrm{ij}}-\mathrm{\&delta;},0\}$

Preferably, adopt roulette wheel dish selection algorithm in described genetic algorithm and retain the optimum system of selection combined, find the chromosome that in population, maximum adaptation degree is corresponding, and with the minimum chromosome of replacement fitness and first chromosome of current population.

Preferably, following self adaptation intersection, mutation operator is adopted in described genetic algorithm:

$P_c\left(C_i\right)=k_c\frac{1-P_s\left(C_i\right)}{t}$

$P_m\left(C_i\right)=k_m\frac{1-P_s\left(C_i\right)}{t}$

Wherein, P _c(C _i), P _m(C _i) be respectively chromosome C _icrossover probability, mutation probability, P _s(C _i) be chromosome C _iselect probability, k _mthe coefficient of variation, k _cbe interaction coefficent, t is current evolutionary generation.

Compared to existing technology, the present invention effectively can reduce the network traffics that monitoring produces, the load balancing of feasible region administrative center, and monitoring time is shorter.

Accompanying drawing explanation

Fig. 1 is network configuration example;

Fig. 2 is chromosome coding example;

Fig. 3 is self-adapted genetic algorithm flow chart;

Fig. 4 is selection opertor flow chart;

Fig. 5 is self adaptation crossover operator flow chart;

Fig. 6 is adaptive mutation rate flow chart.

Embodiment

Below in conjunction with accompanying drawing, technical scheme of the present invention is described in detail:

For a large scale network, an available non-directed graph G=(V, E) represents whole network environment, wherein gathers V={v ₀, v ₁..., v _nrepresent set of node, v ₀represent cloud administrative center; Set V _a( 1<=a<n) the set of representative domain administrative center.E={e _ijrepresent the set of link, wherein, e _ij=(v _i, v _j) represent node v _iand v _jbetween link.Link delay in network is symmetrical, stackable.With function D (e _ij) representative edge e _ijdelay, comprising queueing delay, transmission delay and propagation delay.If limit collection E _ij(Path (v _i, v _j)), represent from node v _ito node v _jshortest path on the set on all limits, then node v _iwith v _jbetween delay be defined as if use Load _irepresent the load of the i-th Ge Yu administrative center, p represents the actual number of network Zhong Yu administrative center, and the load balancing factor lambda of Ze Yu administrative center is defined as:

$\mathrm{\&lambda;}=\sqrt{\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{|{\mathrm{Load}}_i-\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{Load}}_i}{p}|}^2}{p-1}}---\left(1\right)$

Following Mathematical Modeling can be set up for the problem of Region dividing in stratification monitoring system:

$\mathrm{Minf}(x_{\mathrm{ij}},y_i)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}T{f}_{\mathrm{ij}}{x}_{\mathrm{ij}}+\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i(\mathrm{rtT}{f}_j+2\mathrm{MaTf})---\left(2\right)$

s.t.

$\underset{j=1}{\overset{\left|V\right|}{\mathrm{\&Sigma;}}}{x}_{\mathrm{ij}}=1,(\&ForAll;v_i\&Element;V_a)---\left(3\right)$

$\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_{\mathrm{ij}}\&le;y_j,j=\mathrm{1,2}...,n---\left(4\right)$

$\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i\&le;P---\left(5\right)$

$\underset{e_{\mathrm{ij}}\&Element;E}{\mathrm{\&Sigma;}}{b}_e^{\mathrm{ij}}D\left(e_{\mathrm{ij}}\right)x_{\mathrm{ij}}\&le;\mathrm{\&delta;},i,j=\mathrm{1,2}...,n---\left(6\right)$

λ<Bthreshold (7)

Wherein:

V{v ₀, v ₁..., v _nrepresent the set of all nodes in network, v ₀represent network monitoring center, n represents monitored network node number; Set V _a( 1<=a<n) the set of representative domain administrative center.E={e _ijrepresent the set of all links in network, wherein, e _ij=(v _i, v _j) represent node v _iand v _jbetween link;

Tf _ijrepresentative domain administrative center v _jmonitor its management domain interior joint v _itime produce flow;

RtTf _jrepresentative domain administrative center v _jby the flow produced when sending it back Surveillance center after the Information procession of collection;

MaTf represents the monitoring flow that any one node of monitoring is once produced;

P represents the maximum number selecting territory administrative center;

D (e _ij) represent link e _ijdelay;

δ is delay-tolerant coefficient;

λ is the load balancing factor of territory administrative center;

BThreshold represents the threshold value of the load balancing factor.

Above-mentioned various implication: the target function of formula (2) represents that the flow that network monitor produces is minimum, and namely occupied bandwidth minimizes.The constraints of formula (3) represents monitoring node v _ijust a Ge Yu administrative center is distributed to.Formula (4) represents in a territory to only have a Ge Yu administrative center.The number of formula (5) representative domain administrative center is no more than P.Formula (6) is that the monitoring information ensureing between territory administrative center node and monitored node is assembled and met deferred constraint, reduces monitoring time as far as possible.Formula (7) retrains the load balancing of each territory administrative center.

Above-mentioned Mathematical Modeling is optimized and solves, such as, adopt greedy algorithm, Lagrangian Arithmetic, Means of Penalty Function Methods, particle cluster algorithm etc., monitored area splitting scheme can be obtained.For this Mathematical Modeling, the present invention adopts following methods to be optimized and solves: first utilize genetic algorithm to determine the position of each territory administrative center, namely select the node as territory administrative center; Then, other peer distribution in network are given according to delay optimal policy the territory administrative center chosen, whole network is divided into multiple monitored area.

Genetic algorithm uses for reference a kind of global optimizing algorithm of living nature natural selection and evolution.The solution of problem is encoded into individuality (also known as chromosome) by it, multiple individuality forms initial population, based on the natural selection mechanism of the survival of the fittest, utilize the operators such as selection, intersection, variation that initial population is constantly evolved, direction towards optimal solution is moved, and finally searches the optimal solution of problem.

Postponing optimal policy is decide the ownership of node to the delay of territory administrative center according to network node.First, computing node is to the delay of each territory administrative center, and manage then to by this peer distribution the minimum territory administrative center of delay, this strategy can ensure the real-time of monitoring.

Optimum derivation algorithm of the present invention specifically comprises following content:

1. the initialization of chromosomal coding and population

In genetic algorithm, first separate { y by one of above-mentioned model _jbe mapped to a chromosome, y _jvalue is mapped as a chromosomal gene code; Using the sequence number of node as chromosomal gene order number, using the value of a chromosome jth gene order as y _j, namely gene position is 1, represents that node corresponding to this gene sequence number is chosen as territory administrative center; Be 0, represent that node corresponding to this gene sequence number is not selected as territory administrative center.

Next, the ordinary node not being selected as territory administrative center is distributed to territory administrative center and form a monitored area splitting scheme, namely according to { y _jgenerating solution { x _ij.The present invention is according to y _j=1 territory administrative center chosen, will remain peer distribution to these territory administrative centers according to the strategy of time delay optimum, form the solution { x of a Region dividing _ij, so-called time delay optimal policy, is about to residue peer distribution and manages to the territory administrative center the shortest with its time delay.

Setting initial population scale M, maximum evolutionary generation T, according to certain constraints, a stochastic generation M chromosome is as initial population P (0).

2. the design of fitness function

Need to design corresponding fitness function to evaluate quality individual in population.Fitness function value is less, represents chromosomal adaptedness better.Otherwise, then poorer.By the individuality that fitness function is selected, should ensure that monitoring on the basis meeting deferred constraint, can reduce by the caused flow of monitoring itself as far as possible, and ensure the load balancing of each territory administrative center.Design fitness function is as follows:

$\mathrm{fit}\left(\mathrm{Ci}\right)=\frac{\mathrm{CTf}\left(\mathrm{Ci}\right)+\mathrm{NTf}\left(\mathrm{Ci}\right)}{\mathrm{MaxTf}}+\frac{\mathrm{\&lambda;}\left(\mathrm{Ci}\right)}{\mathrm{BThreshold}}+\mathrm{Overdelay}\left(\mathrm{Ci}\right)---\left(8\right)$

Wherein, Ci represents a chromosome, and corresponding area division scheme is ({ x _ij, { y _j); CTf(Ci) represent under the area division scheme that chromosome Ci is corresponding, the total flow produced when territory administrative center and administrative center are mutual, comprise data transfer throughput with monitor produce flow, its computing formula is as follows:

$\mathrm{CTf}\left(\mathrm{Ci}\right)=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i(\mathrm{rtT}{f}_j+\mathrm{Matf})---\left(9\right)$

The total flow produced when each node is mutual in NTf (Ci) representative domain administrative center and its management domain, comprise transmitting data stream amount equally and monitor the flow sum produced, its computing formula is as follows:

$\mathrm{NTf}\left(\mathrm{Ci}\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_{\mathrm{ij}}T{f}_{\mathrm{ij}}+y_j\mathrm{Matf})---\left(10\right)$

MaxTf represents when not zoning, the total flow that monitoring produces, and its value is that all nodes and administrative center directly transmit flow that data produce and monitor flow, and its computing formula is as follows:

$\mathrm{MaxTf}=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}(T{f}_{0j}+\mathrm{Matf})---\left(11\right)$

λ (Ci) is under the corresponding scheme of Ci, the load balancing factor of territory administrative center;

Overdelay (Ci) is the deferred constraint factor, if when territory administrative center is less than δ (delay-tolerant coefficient) to the delay of management domain interior joint, show that current network delay is rational, acceptable, the calculating of reply fitness does not have an impact, so the value arranging Overdelay is 0.Once territory administrative center postpones to be greater than δ to management domain interior joint, represent and postpone to exceed tolerance interval, its value is larger, then fitness function value should be larger.Therefore, its computing formula can be expressed as:

$\mathrm{Overdelay}\left(\mathrm{Ci}\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i\max \{\underset{e_{\mathrm{ij}}\&Element;E}{\mathrm{\&Sigma;}}{b}_e^{\mathrm{ij}}D\left(e_{\mathrm{ij}}\right)x_{\mathrm{ij}}-\mathrm{\&delta;},0\}---\left(12\right)$

3. the design of selection opertor

What the genetic algorithm of standard adopted is roulette selection algorithm, due to the randomness of its operation, so the error selected is larger.In order to ensure that the individuality of fitness optimum can be selected as far as possible, the present invention adopts roulette wheel dish selection algorithm and retains the optimum method combined: find the chromosome that in population, maximum adaptation degree is corresponding, and with the minimum chromosome of replacement fitness and first chromosome of current population, ensure the heredity of optimum individual as far as possible.Wherein, the computational methods of select probability employing roulette are:

$P_s\left(C_i\right)=\mathrm{fit}\left(C_i\right)/\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}\mathrm{fit}\left(C_j\right)---\left(13\right)$

Wherein, fit (C _i) be chromosome C _ifitness, M is chromosome number in population.

4. the design of adaptive intersection, mutation operator

In order to ensure the diversity of computing initial stage population, the later stage can Fast Convergent, intersects, mutation probability constantly should reduce along with the increase of evolutionary generation.Meanwhile, if increase the low chromosomal crossover and mutation probability of fitness, reduce the high chromosomal variation of fitness, crossover probability, then can optimize population while increase chromosome multiformity.Therefore, the present invention devises a kind of self adaptation intersection, mutation operator, changes the size of intersection, mutation probability, that is: according to chromosomal fitness and evolutionary generation

$P_c\left(C_i\right)=k_c\frac{1-P_s\left(C_i\right)}{t}---\left(14\right)$

$P_m\left(C_i\right)=k_m\frac{1-P_s\left(C_i\right)}{t}---\left(15\right)$

P _c(C _i), P _m(C _i) be respectively chromosome C _icrossover probability, mutation probability, P _s(C _i) be chromosome C _iselect probability, k _cinteraction coefficent, k _mbe the coefficient of variation, t is current evolutionary generation.

Formula (14) and formula (15) show, at the algorithm initial stage, the probability of crossover and mutation is comparatively large, improves population diversity, adds the new individual ability of algorithm detection, overcomes Premature Convergence; In the algorithm later stage, crossover probability and mutation probability reduce, and add convergence of algorithm speed.In the same generation population, algorithm also uses different crossover and mutation probability for different individualities, the individuality larger to fitness value, and crossover and mutation probability is less, and the advantage remained in population is individual; And the individuality that appropriateness value is less, crossover and mutation probability is larger, accelerates swarm optimization.

For the ease of public understanding technical solution of the present invention, carry out Region dividing for the network configuration shown in accompanying drawing 1, the number n=15 of its interior joint, node 0 represents supervision and management center, other nodes are ordinary node, and the delay between node is between [0.1-04] ms.Monitored area of the present invention divides and first adopts genetic algorithm to determine the position of each territory administrative center, and algorithm flow as shown in Figure 3, specifically comprises the following steps:

1. first encode and initialization of population to the result of Region dividing, detail is as follows:

A) first encode to the Region dividing of whole network, as shown in Figure 2, wherein, each gene position in chromosome represents a node in network to chromosome coding, the ID corresponding node numbering in a network of gene position.Calculate the higher limit P=0.1 × n=1.5 of the number of regional center according to the interstitial content of network, rounding is 2 regional center nodes, then chromosomal gene position be 1 number can not more than 2.

B) initial population scale M=30 is set, maximum evolutionary generation T=100.

C) adopt random function stochastic generation M chromosome, each chromogene position be 1 number be no more than 2.

2. calculate all chromosomal fitness values in population, concrete steps are as follows:

A) for a chromosome, be set to Ci, judge the position at regional center, other nodes in network being distributed to regional center according to postponing optimum strategy, forming an area division scheme ({ x _ij, { y _j).Chromosome as shown in Figure 2, node 1 and node 8 are chosen as district management center, other nodes are distributed to these two nodes according to delay optimal policy and manages, then network is divided into { 8,9,10,11,12,13,14} and { 1,2,3,4,5,6,7} two set, is managed by node 8 and node 1 respectively.

B) according to this zone scheme, the value of the fitness of this area division scheme homologue Ci is calculated by fitness function fit (Ci) in formula (8).

C) step is repeated a), b) to all chromosomes in population, calculate each chromosomal fitness value.

3. Selecting operation, as shown in Figure 4, concrete operations are as follows for its flow process:

A) chromosome that in selected population, fitness is the highest, as first chromosome of population of future generation, and replaces the chromosome that in original seed group, fitness is minimum.

B) each chromosomal select probability Ps in population is calculated according to formula (13).

C) Stochastic choice chromosome C _i, and the random number A between stochastic generation one [0,1].

If d) Ps (C _i) >A, then this chromosome is selected, otherwise gives up.

4. crossing operation, as shown in Figure 5, concrete operations are as follows for its flow process:

A) from population, two chromosome C are selected according to selection opertor _iand C _j.

B) these two chromosomal crossover probability Pc (C are calculated according to formula (14) _i), Pc (C _j).

C) the random number B between stochastic generation one [0,1].

If d) (Pc (C _i)+Pc (C _j))/2>B, then these two chromosomes carry out two-point crossover, new individuality is added population of future generation, otherwise directly puts into population of future generation.

5. mutation operator, as shown in Figure 6, concrete operations are as follows for its flow process:

A) from population, a chromosome C is selected according to selection opertor _i.

B) this chromosomal mutation probability Pm (C is calculated according to formula (15) _i).

C) the random number C between stochastic generation one [0,1].

If d) Pm (C _i) >C, then add population of future generation after mutation operation being carried out to the gene position of in this chromosome, otherwise directly put into population of future generation.

6. screen, concrete operations are as follows:

A) by gene position in new population be 1 the figure place chromosome that is greater than P give up.

If b) number of new population does not reach M, then repeat 4,5 continue to add.

7. stop judging: if t=T, then the maximum adaptation degree obtained in evolutionary process is individual to be exported as optimal solution, stops calculating; If do not meet and return step 2.

Due in above-mentioned algorithm when calculating chromosome fitness, defining area division scheme according to postponing optimum strategy, therefore, directly choosing the area division scheme corresponding to the maximum chromosome of fitness.

Claims (5)

1. the monitored area division methods of a stratification monitoring network, described stratification monitoring network is divided into multiple monitored area, each monitored area comprises a Ge Yu administrative center, information interaction is carried out for other nodes of monitoring in this monitored area with network monitoring center, it is characterized in that, the splitting scheme of described monitored area is obtained by the following Mathematical Modeling of Optimization Solution:

Min f(x _ij,y _j)

$f(x_{\mathrm{ij}},y_j)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}T{f}_{\mathrm{ij}}{x}_{\mathrm{ij}}+\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_j(\mathrm{rt}{\mathrm{Tf}}_j+2\mathrm{MaTf})$

s.t.

$\underset{j=1}{\overset{\left|V\right|}{\mathrm{\&Sigma;}}}{x}_{\mathrm{ij}}=1,(\&ForAll;v_i\&Element;V_a)$

x _ij≤y _j,j＝1,2...,n

$\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_j\&le;P$

$\underset{e_{\mathrm{ij}}\&Element;E}{\mathrm{\&Sigma;}}{b}_e^{\mathrm{ij}}D\left(e_{\mathrm{ij}}\right)x_{\mathrm{ij}}\&le;\mathrm{\&delta;},i,j=\mathrm{1,2}...,n$

λ<BThreshold

Wherein:

V={v ₀, v ₁..., v _nrepresent the set of all nodes in network, v ₀represent network monitoring center, n represents monitored network node number; Set the set of representative domain administrative center; E={e _ijrepresent the set of all links in network, wherein, e _ij=(v _i, v _j) represent node v _iand v _jbetween link;

Tf _ijrepresentative domain administrative center v _jmonitor its management domain interior joint v _itime produce flow;

RtTf _jrepresentative domain administrative center v _jby the flow produced when sending it back Surveillance center after the Information procession of collection;

MaTf represents the monitoring flow that monitoring node is once produced;

P represents the maximum number selecting territory administrative center;

D (e _ij) represent link e _ijdelay;

δ is delay-tolerant coefficient;

λ is the load balancing factor of territory administrative center, is defined as:

$\mathrm{\&lambda;}=\sqrt{\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{|{\mathrm{Load}}_i-\frac{\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{\mathrm{Load}}_i}{p}|}^2}{p-1}};$

Wherein, Load _irepresent the load of the i-th Ge Yu administrative center, the actual number of p representative domain administrative center;

BThreshold represents the threshold value of the load balancing factor.

2. the monitored area division methods of stratification monitoring network as claimed in claim 1, is characterized in that, to described Mathematical Modeling be optimized solve time, first utilize genetic algorithm to determine the position of each territory administrative center; Then, other peer distribution in network are given according to delay optimal policy the territory administrative center chosen, whole network is divided into multiple monitored area.

3. the monitored area division methods of stratification monitoring network as claimed in claim 2, is characterized in that, when utilizing genetic algorithm to determine the position of each territory administrative center, its mapping relations are as follows:

Each chromosome represents a kind of area division scheme, a solution of corresponding described Mathematical Modeling; Be that each node in network composes one No. ID, as chromosomal gene order number with continuous print integer; Chromosomal length is monitored network node number n; Chromosome is encoded in a binary fashion, and the value of gene is 1, represents that node corresponding to this gene sequence number is chosen as territory administrative center;

In genetic algorithm, evaluate each chromosomal fitness function as follows:

$\mathrm{fit}\left(\mathrm{Ci}\right)=\frac{\mathrm{CTf}\left(\mathrm{Ci}\right)+\mathrm{NTf}\left(\mathrm{Ci}\right)}{\mathrm{MaxTf}}+\frac{\mathrm{\&lambda;}\left(\mathrm{Ci}\right)}{\mathrm{BThreshold}}+\mathrm{Overdelay}\left(\mathrm{Ci}\right)$

Wherein, Ci represents a chromosome, and corresponding area division scheme is ({ x _ij, { y _j), CTf (Ci) represents under the area division scheme that chromosome Ci is corresponding, territory administrative center and network monitoring center mutual time the total flow that produces; Under NTf (Ci) represents the zone scheme that Ci is corresponding, the total flow produced when each node is mutual in territory administrative center and its management domain; MaxTf represents when not zoning, the total flow that monitoring produces; λ (Ci) is under the zone scheme that Ci is corresponding, the load balancing factor of territory administrative center; Overdelay (Ci) is the deferred constraint factor, and its computing formula can be expressed as:

$\mathrm{Overdelay}\left(\mathrm{Ci}\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_j\max \{\underset{e_{\mathrm{ij}}\&Element;E}{\mathrm{\&Sigma;}}{b}_e^{\mathrm{ij}}D\left(e_{\mathrm{ij}}\right)x_{\mathrm{ij}}-\mathrm{\&delta;},0\}.$

4. the monitored area division methods of stratification monitoring network as claimed in claim 3, it is characterized in that, adopt roulette wheel dish selection algorithm in described genetic algorithm and retain the optimum system of selection combined: find the chromosome that in population, maximum adaptation degree is corresponding, and with the minimum chromosome of replacement fitness and first chromosome of current population.

5. the monitored area division methods of stratification monitoring network as claimed in claim 3, is characterized in that, adopts following self adaptation intersection, mutation operator in described genetic algorithm:

$P_c\left(C_i\right)=k_c\frac{1-P_s\left(C_i\right)}{t}$

$P_m\left(C_i\right)=k_m\frac{1-P_s\left(C_i\right)}{t}$

Wherein, P _c(C _i), P _m(C _i) be respectively chromosome C _icrossover probability, mutation probability, P _s(C _i) be chromosome C _iselect probability, k _mthe coefficient of variation, k _cbe interaction coefficent, t is current evolutionary generation.