CN103354545B - Cloud computing based P2P streaming media server cluster deploying method - Google Patents

Abstract

The present invention provides a cloud computing based P2P streaming media server cluster deploying method, in which the bandwidth requirement of a P2P streaming media live broadcast system with delay protection can be predicted, and a virtual streaming media server cluster is built on a cloud platform DCN network. Specially, a method for predicting the bandwidth requirement of the P2P streaming media live broadcast system with delay protection is based on data package delay protection, and has parameters of delay protection time, and push-pull percentage and uplink bandwidth available rate of a P2P topology network, thereby building a bandwidth requirement model for the P2P streaming media live broadcast system for bandwidth prediction. A method for building the P2P streaming media server cluster on the cloud platform DCN network is based on the bandwidth requirement of the P2P streaming media live broadcast system with delay protection, and builds a logical structure for the P2P streaming media server cluster. The P2P streaming media server cluster is deployed on a cloud platform by constructing a deployment matrix of the P2P streaming media server cluster.

Description

A kind of P2P streaming media server clustered deploy(ment) method based on cloud computing

Technical field

The present invention relates to technical field of computer multimedia, relate to a kind of P2P streaming media server clustered deploy(ment) method based on cloud computing, prediction has the bandwidth demand of the P2P flow medium live system that time delay ensures, thus draws the P2P streaming media server clustered deploy(ment) strategy on cloud platform DCN network.

Background technology

Along with Internet technology application increasingly extensive, net cast service has become the important component part of people's daily life, and the quality of its service quality largely depends on that prediction has bandwidth demand and the P2P streaming media server cluster structure of the P2P flow medium live system that time delay ensures.Applicant is new through looking into, and retrieves 3 sections of related to the present invention patents, they respectively:

1. can prevent the P2P stream media live broadcasting method [application number: 200910037247.X] of flash crowd

The topological construction method [application number: 200910088670.2] of 2.P2P stream media network

3. the FBSTREAM live streaming media model [application number: 200810028299.6] based on P2P network

Above-mentioned existing patent Problems existing is respectively: can prevent from the P2P stream media live broadcasting method of flash crowd from adopting increasing service unit carrying out optimum allocation, introduce content distributing server and directly provide service for user, alleviate flash crowd phenomenon, can not fundamentally solve delay congestion problems; The topological construction method of P2P stream media network is all, based on the loading condition of single server, balanced distribution is carried out in the request of user, and does not consider the Function and operation between cluster server; A kind of FBSTREAM live streaming media model based on P2P network does not consider the minimum requirements that active user postpones and the available minimum bandwidth of server.

Summary of the invention

In order to solve the deficiency about stream media server system and stream media live broadcasting method patent in background technology, the object of the invention is to propose a kind of P2P streaming media server clustered deploy(ment) method based on cloud computing, with packet delay guarantee for foundation, ensure that in time, P2P topological network, push-and-pull percentage, upstream bandwidth provide rate for parameter with time delay, build P2P flow medium live system bandwidth demand model prediction bandwidth; Build and have with the bandwidth demand of the P2P flow medium live system of time delay guarantee for foundation, structure P2P streaming media server clustering logic structure, and build P2P streaming media server clustered deploy(ment) matrix, on P2P streaming media server clustered deploy(ment) to cloud platform.

For reaching above object, the present invention takes following technical scheme to be achieved:

Based on a P2P streaming media server clustered deploy(ment) method for cloud computing, comprise following content: build P2P flow medium live system bandwidth demand model, to predict the bandwidth demand with the P2P flow medium live system that time delay ensures; With the bandwidth demand of prediction for foundation, build P2P streaming media server clustering logic structure; With P2P streaming media server clustering logic structure for foundation, build P2P streaming media server communication cost Matrix C _p2p; With the switch number of process between cloud platform DCN network virtual machine for foundation, build cloud platform DCN network cost Matrix C _v; With C _p2pand C _vfor foundation, build P2P streaming media server clustered deploy(ment) matrix, and then by P2P streaming media server clustered deploy(ment) to cloud platform.

The concrete steps of described structure P2P flow medium live system bandwidth demand model are:

Step1: initiation parameter: propelling movement multiple is α=1, subflow pushes speed and reduces proportion omegab=0.4;

Step2: utilize formula S _pac=C ^(h)r/s calculates single times of subflow and pushes generation service packs speed, and wherein r is Streaming Media speed, S _paxfor patching stream speed, s is subflow number, C ^(h)for path tree adds up the logic child node quantity that can produce patch request from 0 to h layer;

Step3: if S _pacω>=α r, then double propelling movement multiple α; Otherwise jump to Step5;

Step4: make S _pac=S _pac(1-ω), jumps to Step3;

Step5: server bandwidth load is then S=α r+S _pac(S _pacω>=α r), wherein S represents server demands bandwidth.

Described path tree adds up the logic child node quantity C that can produce patch request from 0 to h layer ^(h)following formula is utilized to calculate:

Wherein $f(h,n)=C_h^0{p}^0{q}^h+C_h^1{p}^1{q}^{h-1}+\·\·\·+C_h^n{p}^h{q}^{h-n},$ P is for pushing away packet proportion, and q, for drawing packet proportion, meets p+q=1; Wherein k ^(h)=ρ k ^(h-1) cr/r=ρ ck ^(h-1), it is the upstream bandwidth value of disposing local real network according to P2P system that ρ gets empirical value 0.98, c, k ^(h-1)for by k ⁽⁰⁾recursion value, k ⁽⁰⁾equal subflow quantity, the number of plies of h path tree, γ is the ratio drawing packet delay Yu push away packet delay, and d is delay guaranteed time and the ratio drawing packet delay.

The concrete steps of described structure P2P streaming media server clustering logic structure are as follows:

Step1: obtaining bandwidth load demand according to the P2P flow medium live system bandwidth demand model built is S;

Step2: note D _n[0-n] is the list of root P2P streaming media server, and D [0-n] is P2P streaming media server list to be added, and P [0-n] is the P2P streaming media server list disposed, and does not comprise root P2P streaming media server; Usage degree is that unit represents P2P streaming media server remaining bandwidth, the Streaming Media speed of each degree expression one times;

Step3: by D [n] according to order sequence from big to small, get server D [k] in D [0-n] successively, access root server D _nin cluster multicast tree under [k];

Step4: for D [k] selects father's node: if D _n[k] spends non-vanishing, then be directly connected on D _nunder [k], by D _n[k] degree subtracts one, if D _n[k] degree is zero, the server access selecting redundancy maximum from P [0-n] according to breadth First principle, and father's node degree subtracts one, is incorporated to by D [k] in P [0-n];

Step5: judge that cluster multicast sets the bandwidth that can provide, if be more than or equal to S, then terminate, otherwise perform Step3.

Described structure P2P streaming media server communication cost Matrix C _p2pconcrete steps be:

Step1: initialization P2P streaming media server communication cost Matrix C _p2p, make C _p2p(i, j)=0, wherein i, j represent the numbering of P2P streaming media server;

Step2: according to constructed P2P streaming media server clustering logic structure tree, adopts pre-reset mechanism method, travels through whole logic tree; If the forwarding rate that there is Streaming Media between P2P streaming media server i and P2P streaming media server j is v, then remember C _p2p(i, j)=v; Otherwise note C _p2p(i, j)=0;

Step3: obtain complete P2P streaming media server communication cost Matrix C _p2p.

Described structure cloud platform DCN network cost Matrix C _vconcrete steps are:

Step1: initialization cloud platform DCN network cost Matrix C _v, make C _v(i, j)=0, wherein i, j represent the numbering of virtual machine;

Step2: for any two virtual machine i and virtual machine j, if virtual machine i and virtual machine j on same host's physical machine or i equal j, then remember C _v(i, j)=0; Otherwise jump to Step3;

Step3: if host's physical machine of virtual machine i and virtual machine j is connected on same access switch, then remember C _v(i, j)=1; Otherwise jump to Step4;

Step4: if virtual machine i is connected not on same access switch with host's physical machine of virtual machine j, and in same aggregation switch, then remember C _v(i, j)=3; Otherwise jump to Step5;

Step5: remember C in other situations _v(i, j)=5;

Step6: obtain complete cloud platform DCN network cost Matrix C _v.

The concrete steps of described structure P2P streaming media server clustered deploy(ment) matrix are as follows:

Step1: initialization ant group algorithm iterations NC, heuristic function τ _ijbe 1, pheromones η _ij; The initialization of pheromones: by cloud platform DCN network cost Matrix C _vthe matrix A obtaining 1*n is added, by P2P streaming media server communication cost Matrix C according to row _p2pbe added according to row and obtain 1*n matrix B, η _ij=A*B', is placed on the P2P streaming media server of node 0 by r ant;

Step2: in t, ant k moves according to state transition function, migration function meet shown in following formula:

$p_{\mathrm{ij}}^k\left(t\right)=\left\{\begin{array}{cc}\frac{{\left[{\mathrm{\τ}}_{\mathrm{ij}}\left(t\right)\right]}^{\mathrm{\α}}{\left[{\mathrm{\η}}_{\mathrm{ij}}\right]}^{\mathrm{\β}}}{\underset{r\∈{\mathrm{allowed}}_k}{\mathrm{\Σ}}{\left[{\mathrm{\τ}}_{\mathrm{ir}}\left(t\right)\right]}^{\mathrm{\α}}{\left[{\mathrm{\η}}_{\mathrm{ir}}\right]}^{\mathrm{\β}}}& j\∈{\mathrm{allowed}}_k\\ 0& \mathrm{otherwise}\end{array}\right.$

In formula,

Allowed _krepresent the set of the deployment point that current ant can be selected;

α value is 1, and represent that heuristic function selects deployment point Decision Making Effect ant, β value is 5, represents that pheromones selects deployment point Decision Making Effect ant;

Step3: local updating pheromones value η _ij, the value to every paths lastest imformation element that ant is passed by:

Step4: if all r ant all constructs solution, then turn to step6, otherwise forward step2 to;

Step5: if exceed iterative cycles times N C, then forward step6 to; Otherwise forward step2 to;

Step6: ant has selected the n bar arc between n P2P streaming media server and n virtual machine, and n bar arc can construct deployment matrix thus.

Compared with prior art, advantage of the present invention is:

1, virtual stream media services number of clusters is the P2P flow medium live system with delay guarantee according to using push-and-pull hybrid protocol, and in prediction P2P topological network, the server bandwidth of demand obtains.

If when 2, P2P node exceedes maximum delay restriction, can directly to server request data, be also service packs, this has fundamentally ensured delay.

3, consider to minimize to take the cloud platform DCN network bandwidth, by the clustered deploy(ment) of P2P streaming media server on cloud platform, between Deterministic service device, bandwidth is minimum, alleviates server cluster burden.

Accompanying drawing explanation

Fig. 1 is a kind of P2P streaming media server clustered deploy(ment) method system Organization Chart based on cloud computing of the present invention.

Fig. 2 is P2P flow medium live system bandwidth demand model step figure of the present invention.

Fig. 3 is structure P2P streaming media server clustering logic configuration steps figure of the present invention.

Fig. 4 is that the present invention builds P2P streaming media server communication cost Matrix C _vblock diagram.

Fig. 5 is that the present invention builds cloud platform DCN network cost Matrix C _vblock diagram.

Fig. 6 is the block diagram that the present invention builds P2P streaming media server clustered deploy(ment) matrix.

Embodiment

Describe the present invention below in conjunction with drawings and Examples.

System architecture

With reference to shown in Fig. 1, the present invention is based on the framework of the P2P streaming media server clustered deploy(ment) method of cloud computing, be made up of P2P registrar node, virtual stream media server cluster and P2P network node, specific as follows:

P2P registrar node: for monitoring the state of virtual stream media server, controls unlatching and the shutoff operation of virtual stream media server;

Virtual stream media server cluster: the server specifically carrying out direct broadcast service to user; Virtual stream media services cluster is constituted by multiple virtual stream media server;

P2P network node: send data packets to each other between P2P network node, packet in P2P network through the push-and-pull data of multi-hop, if node accepts packet delay be greater than delay guarantee, then it directly sends request to server cluster, and backward its of server cluster response propagates video/audio.

P2P streaming media server bandwidth load modeling strategy.

For the P2P flow medium live system with delay guarantee using push-and-pull hybrid protocol, the server bandwidth of demand in prediction P2P topological network.Due to the push-and-pull data of multi-hop in P2P network, cause packet delay increasing.Although what formed between each node and each node in P2P network is netted structure, but experimentally find the diffusion of P2P topological network packet, be actually and carrying out transmitting in stable, implicant mode, the approach claiming packet diffusion is subflow number, consider this point, state the service packs quantity that each layer produces formally.Spread with the form of stable implicit tree in P2P network according to packet, formalization representation service packs quantity.With reference to shown in Fig. 2, this operation implementation procedure is as follows:

Step1: initiation parameter: propelling movement multiple is α=1, subflow pushes speed and reduces proportion omegab=0.4;

Step2: utilize formula S _pac=C ^(h)r/s calculates single times of subflow and pushes generation service packs speed, and wherein r is Streaming Media speed, S _pacfor patching stream speed, s is subflow number, C ^(h)for path tree adds up the logic child node quantity that can produce patch request from 0 to h layer;

Step3: if S _pacω>=α r, then double propelling movement multiple α; Otherwise jump to Step5;

Step4: make S _pac=S _pac(1-ω), jumps to Step3;

Step5: server bandwidth load is then S=α r+S _pac(S _pacω>=α r), wherein S represents server demands bandwidth.

Based on cloud computing virtualized P2P streaming media server clustering logic structure strategy

Build P2P streaming media server logical construction, with reference to Fig. 3, this operation implementation procedure is as follows:

Step1: obtaining bandwidth load demand according to the P2P flow medium live system bandwidth demand model built is S;

Step2: note D _n[0-n] is the list of root P2P streaming media server, and D [0-n] is P2P streaming media server list to be added, and P [0-n] is the P2P streaming media server list disposed, and does not comprise root P2P streaming media server; Usage degree is that unit represents P2P streaming media server remaining bandwidth, the Streaming Media speed of each degree expression one times;

Step3: by D [n] according to order sequence from big to small, get server D [k] in D [0-n] successively, access root server D _nin cluster multicast tree under [k];

Step4: for D [k] selects father's node: if D _n[k] spends non-vanishing, then be directly connected on D _nunder [k], by D _n[k] degree subtracts one, if D _n[k] degree is zero, the server access selecting redundancy maximum from P [0-n] according to breadth First principle, and father's node degree subtracts one, is incorporated to by D [k] in P [0-n];

Step5: judge that cluster multicast sets the bandwidth that can provide, if be more than or equal to S, then terminate, otherwise perform Step3.

Build P2P streaming media server communication cost matrix strategy

Build P2P server communication cost matrix C _p2pserver cluster logical construction according to step1 builds: if there is the forwarding rate v of stream medium data between server i and server j, then remember C _p2p(i, j)=v; If there is not Streaming Media speed between server i and server j, be designated as C _p2p(i, j)=0.With reference to Fig. 4, this operation implementation procedure is as follows:

Step1: initialization P2P streaming media server communication cost Matrix C _p2p, make C _p2p(i, j)=0, wherein i, j represent the numbering of P2P streaming media server;

Step2: the P2P streaming media server clustering logic structure tree constructed by claim 4, adopts pre-reset mechanism method, travels through whole logic tree; If the forwarding rate that there is Streaming Media between P2P streaming media server i and P2P streaming media server j is v, then remember C _p2p(i, j)=v; Otherwise note C _p2p(i, j)=0;

Step3: obtain complete P2P streaming media server communication cost Matrix C _p2p.

Build cloud platform DCN network cost matrix strategy

Build the cost matrix between cloud platform DCN network virtual server, set up the cost of cloud platform DCN network TREE, VL2, Fat-Tree, Bcube equal matrix of different framework.For Tree framework: if two-server is under same access switch, note cost is 1; If two-server is not under same access switch, and under same aggregation switch, then cost is 3; In all the other situations, cost is 5.With reference to Fig. 5, this operation implementation procedure is as follows:

Step1: initialization cloud platform DCN network cost Matrix C _v, make C _v(i, j)=0, wherein i, j represent the numbering of virtual machine;

Step2: for any two virtual machine i and virtual machine j, if virtual machine i and virtual machine j on same host's physical machine or i equal j, then remember C _v(i, j)=0; Otherwise jump to Step3;

Step3: if host's physical machine of virtual machine i and virtual machine j is connected on same access switch, then remember C _v(i, j)=1; Otherwise jump to Step4;

Step4: if virtual machine i is connected not on same access switch with host's physical machine of virtual machine j, and in same aggregation switch, then remember C _v(i, j)=3; Otherwise jump to Step5;

Step5: remember C in other situations _v(i, j)=5;

Step6: obtain complete cloud platform DCN network cost Matrix C _v.

Build P2P streaming media server clustered deploy(ment) matrix

According to cloud platform DCN network cost matrix and P2P streaming media server communication cost matrix, on P2P server disposition to cloud platform DCN virtual server.When making on P2P server disposition to cloud platform virtual server, the bandwidth of consumption is minimum.Take ant group algorithm, with reference to Fig. 6, this operation implementation procedure is as follows:

Step1: initialization ant group algorithm iterations NC, heuristic function τ _ijbe 1, pheromones η _ij.The initialization of pheromones: Matrix C (i, j) is added the matrix A obtaining 1*n according to row, is added D (i, j) according to row and obtains 1*n matrix B, η _ij=A*B'.R ant is placed on the P2P streaming media server of node 0;

Step2: in t, ant k moves according to state transition function, migration function meet shown in following formula:

$p_{\mathrm{ij}}^k\left(t\right)=\left\{\begin{array}{cc}\frac{{\left[{\mathrm{\τ}}_{\mathrm{ij}}\left(t\right)\right]}^{\mathrm{\α}}{\left[{\mathrm{\η}}_{\mathrm{ij}}\right]}^{\mathrm{\β}}}{\underset{r\∈{\mathrm{allowed}}_k}{\mathrm{\Σ}}{\left[{\mathrm{\τ}}_{\mathrm{ir}}\left(t\right)\right]}^{\mathrm{\α}}{\left[{\mathrm{\η}}_{\mathrm{ir}}\right]}^{\mathrm{\β}}}& j\∈{\mathrm{allowed}}_k\\ 0& \mathrm{otherwise}\end{array}\right.$

In formula,

Allowed _krepresent the set of the deployment point that current ant can be selected;

α value is 1, and represent that heuristic function selects deployment point Decision Making Effect ant, β value is 5, represents that pheromones selects deployment point Decision Making Effect ant.

Step3: local updating pheromones value η _ij.Value to the routing update pheromones that every bar ant is passed by:

Step4: if all r ant all constructs solution, then turn to step6, otherwise forward step2 to;

Step5: if exceed iterative cycles times N C, then forward step6 to; Otherwise forward step2 to;

Step6: ant has selected the n bar arc between n P2P streaming media server and n virtual machine, and n bar arc can construct deployment matrix thus.

Claims (1)

1. based on a P2P streaming media server clustered deploy(ment) method for cloud computing, it is characterized in that: build P2P flow medium live system bandwidth demand model, to predict the bandwidth demand with the P2P flow medium live system that time delay ensures; With the bandwidth demand of prediction for foundation, build P2P streaming media server clustering logic structure; With P2P streaming media server clustering logic structure for foundation, build P2P streaming media server communication cost Matrix C _p2p; With the switch number of process between cloud platform DCN network virtual machine for foundation, build cloud platform DCN network cost Matrix C _v; With C _p2pand C _vfor foundation, build P2P streaming media server clustered deploy(ment) matrix, and then by P2P streaming media server clustered deploy(ment) to cloud platform,

The concrete steps of described structure P2P flow medium live system bandwidth demand model are:

Step1.1: initiation parameter: propelling movement multiple is α=1, subflow pushes speed and reduces proportion omegab=0.4;

Step1.2: utilize formula S _pac=C ^(h)r/s calculates single times of subflow and pushes generation service packs speed, and wherein r is Streaming Media speed, S _pacfor patching stream speed, s is subflow number, C ^(h)for path tree adds up the logic child node quantity that can produce patch request from 0 to h layer;

Step1.3: if S _pacω>=α r, then double propelling movement multiple α; Otherwise jump to Step1.5;

Step1.4: make S _pac=s _pac(1-ω), jumps to Step1.3;

Step1.5: server bandwidth load is then S=α r+S _pac(S _pacω>=α r), wherein S represents server demands bandwidth;

Described path tree adds up the logic child node quantity C that can produce patch request from 0 to h layer ^(h)following formula is utilized to calculate:

Wherein $f(h,n)=C_h^0{p}^0{q}^h+C_h^1{p}^1{q}^{h-1}+...+C_h^n{p}^h{q}^{h-n},$ P is for pushing away packet proportion, and q, for drawing packet proportion, meets p+q=1; Wherein k ^(h)=ρ k ^(h-1)cr/r=ρ ck ^(h-1), it is the upstream bandwidth value of disposing local real network according to P2P system that ρ gets empirical value 0.98, c, k ^(h-1)for by k ⁽⁰⁾recursion value, k ⁽⁰⁾equal subflow quantity, the number of plies of h path tree, γ is the ratio drawing packet delay Yu push away packet delay, and d is delay guaranteed time and the ratio drawing packet delay;

The concrete steps of described structure P2P streaming media server clustering logic structure are as follows:

Step2.1: obtaining bandwidth load demand according to the P2P flow medium live system bandwidth demand model built is S;

Step2.2: note D _n[0-n] is the list of root P2P streaming media server, and D [0-n] is P2P streaming media server list to be added, and P [0-n] is the P2P streaming media server list disposed, and does not comprise root P2P streaming media server; Usage degree is that unit represents P2P streaming media server remaining bandwidth, the Streaming Media speed of each degree expression one times;

Step2.3: by D [n] according to order sequence from big to small, get server D [k] in D [0-n] successively, access root server D _nin cluster multicast tree under [k];

Step2.4: for D [k] selects father's node: if D _n[k] spends non-vanishing, then be directly connected on D _nunder [k], by D _n[k] degree subtracts one, if D _n[k] degree is zero, the server access selecting redundancy maximum from P [0-n] according to breadth First principle, and father's node degree subtracts one, is incorporated to by D [k] in P [0-n];

Step2.5: judge that cluster multicast sets the bandwidth that can provide, if be more than or equal to S, then terminate, otherwise perform Step2.3;

Described structure P2P streaming media server communication cost Matrix C _p2pconcrete steps be:

Step3.1: initialization P2P streaming media server communication cost Matrix C _p2p, make C _p2p(i, j)=0, wherein i, j represent the numbering of P2P streaming media server;

Step3.2: according to constructed P2P streaming media server clustering logic structure tree, adopts pre-reset mechanism method, travels through whole logic tree; If the forwarding rate that there is Streaming Media between P2P streaming media server i and P2P streaming media server j is v, then remember C _p2p(i, j)=v; Otherwise note C _p2p(i, j)=0;

Step3.3: obtain complete P2P streaming media server communication cost Matrix C _p2p;

Described structure cloud platform DCN network cost Matrix C _vconcrete steps be:

Step4.1: initialization cloud platform DCN network cost Matrix C _v, make C _v(i, j)=0, wherein i, j represent the numbering of virtual machine;

Step4.2: for any two virtual machine i and virtual machine j, if virtual machine i and virtual machine j on same host's physical machine or i equal j, then remember C _v(i, j)=0; Otherwise jump to Step4.3;

Step4.3: if host's physical machine of virtual machine i and virtual machine j is connected on same access switch, then remember C _v(i, j)=1; Otherwise jump to Step4.4;

Step4.4: if virtual machine i is connected not on same access switch with host's physical machine of virtual machine j, and in same aggregation switch, then remember C _v(i, j)=3; Otherwise jump to Step4.5;

Step4.5: remember C in other situations _v(i, j)=5;

Step4.6: obtain complete cloud platform DCN network cost Matrix C _v;

The concrete steps of described structure P2P streaming media server clustered deploy(ment) matrix are as follows:

Step5.1: initialization ant group algorithm iterations NC, heuristic function τ _ijbe 1, pheromones η _ij; The initialization of pheromones: by cloud platform DCN network cost Matrix C _vthe matrix A obtaining 1*n is added, by P2P streaming media server communication cost Matrix C according to row _p2pbe added according to row and obtain 1*n matrix B, η _ij=A*B', is placed on the P2P streaming media server of node 0 by r ant;

Step5.2: in t, ant k moves according to state transition function, migration function meet shown in following formula:

$p_{\mathrm{ij}}^k\left(k\right)=\left\{\begin{array}{cc}\frac{{\left[{\mathrm{\τ}}_{\mathrm{ij}}\left(t\right)\right]}^{\mathrm{\α}}{\left[{\mathrm{\η}}_{\mathrm{ij}}\right]}^{\mathrm{\β}}}{\underset{r\∈\mathrm{allowe}{d}_k}{\mathrm{\Σ}}{\left[{\mathrm{\τ}}_{\mathrm{ir}}\left(t\right)\right]}^{\mathrm{\α}}{\left[{\mathrm{\η}}_{\mathrm{ir}}\right]}^{\mathrm{\β}}}& j\∈{\mathrm{allowed}}_k\\ 0& \mathrm{otherwise}\end{array}\right.$

In formula,

Allowed _krepresent the set of the deployment point that current ant can be selected;

α value is 1, and represent that heuristic function selects deployment point Decision Making Effect ant, β value is 5, represents that pheromones selects deployment point Decision Making Effect ant;

Step5.3: local updating pheromones value η _ij, the value to every paths lastest imformation element that ant is passed by:

Step5.4: if all r ant all constructs solution, then turn to step5.6, otherwise forward step5.2 to;

Step5.5: if exceed iterative cycles times N C, then forward step5.6 to; Otherwise forward step5.2 to;

Step5.6: ant has selected the n bar arc between n P2P streaming media server and n virtual machine, and n bar arc can construct deployment matrix thus.