CN107172125B - A Fair Bandwidth Allocation Algorithm for Cross-layer P2P Resource Sharing Network - Google Patents

Abstract

A cross-layer P2P resource sharing network bandwidth fair allocation algorithm, each service requester of P2P network calculates the aggregate bandwidth and the price paid to the whole network according to the bandwidth allocated by each service provider; the service requester obtains an actual price to be paid to the service provider based on the price paid to the network and the price charged by the link in the IP network; the service provider obtains the expected price of the service provider according to the actual price paid to the service provider by the service requester; the service provider adjusts the bandwidth allocated to the service requester at the time t +1 according to the bandwidth allocated to each service requester at the time t, the price paid to the service provider by each service requester and the expected price of the service provider; calculating the aggregate flow on the link by the link in the IP network according to the flow passing through the link, and adjusting the price charged at the t +1 moment; each service provider iterates according to the above steps until an optimal point is reached. The invention has the advantages of fair bandwidth allocation, backbone network flow control and the like.

Description

A Fair Bandwidth Allocation Algorithm for Cross-layer P2P Resource Sharing Network

technical field

The invention relates to the technical field of computer networks, in particular to a cross-layer P2P resource sharing network bandwidth fair allocation algorithm based on a price mechanism.

Background technique

A P2P network is a system that realizes resource sharing by integrating various resources such as storage, computing, and files at the edge of the network. Different from the traditional client/server model, P2P adopts the working model of distributed resource sharing. Each node in the network can contribute resources to the entire network, such as providing file sharing and downloading. Therefore, with the increase of nodes in the network, the service capability provided by the network will also increase, and the increase of the system scale can easily meet the needs of users to obtain resources.

In a P2P resource sharing network, a node that requires resources can be served by multiple other nodes, thus overcoming the limitations of centralized servers, greatly improving the efficiency of network resources, and improving the quality of service for users who demand network resources. . At present, there are mature software to realize P2P network resource sharing and downloading, such as BT (BitToiTent), eMuIe (VeryCd), Thunder (Thunder) and other software by providing rich images, files, books and other resources and a good user experience , attracting a huge user group, so as to achieve a wide range of resource sharing.

In a P2P resource sharing network, each node wants to maximize its download bandwidth when obtaining services provided by other nodes, thereby improving its user experience and satisfaction. Therefore, how to fairly distribute the upload bandwidth of the nodes providing resource services among service requesters becomes very important. At present, there are also related literatures that propose bandwidth allocation schemes for P2P resource sharing networks. For example, Chinese Patent Application No. 201510778746.X, published on 2016.03.30, the patent titled "A Bandwidth Allocation Algorithm in P2P File Sharing Network Based on Price Mechanism", designed a kind of price mechanism based P2P file sharing network The bandwidth allocation algorithm is used to realize the fair distribution of the upload bandwidth of the resource provider among the resource demanders. Chinese Patent Application No. 2016100813716, published on 2016.06.29, the patent titled "Utility-optimized P2P file sharing network bandwidth fair allocation algorithm", designed a utility-optimized P2P file sharing network bandwidth fair allocation algorithm, which can realize coupling type and the optimal point of the uncoupled bandwidth resource allocation utility optimization problem.

In terms of academic papers, "Journal of Southeast University (Natural Science Edition)" 2008 No. S1 "A Game Theory-Based Peer-to-Peer Network Bandwidth Allocation Scheme" aims to solve the problem of multiple heterogeneous download nodes in a peer-to-peer network from multiple source nodes. To solve the problem of bandwidth allocation for downloading, a water-filling algorithm and a peer-to-peer network bandwidth allocation scheme capable of accommodating selfish nodes are proposed. "Computer Applications" 2015 No. 6 "Optimized Bandwidth Allocation Strategy in Peer-to-Peer Streaming Media Network Based on Network Coding" Aiming at the application of P2P streaming media system using network coding technology, a node bandwidth resource balancing based on load transfer is proposed. The strategy is to avoid the node overload caused by the randomness of nodes selecting neighbor nodes and requesting bandwidth resources as much as possible. "Computer Applications" 2015 No. 9 "Peer-to-Peer Network Resource Allocation Based on Nash Bargaining" proposes a resource allocation scheme based on Nash bargaining to ensure the minimum service quality of nodes, and the cooperative nodes obtain more resources than non-cooperative nodes. Moreover, it is proved that the greater the relative bargaining power of the nodes in the cooperative game, the more resources the nodes obtain and the greater the benefits.

The above literature basically designs bandwidth allocation algorithms, flow control algorithms and resource management algorithms from the perspective of the application layer of the peer-to-peer network, and does not consider how these bandwidth allocation algorithms are integrated with the existing backbone networks such as IP networks, that is, the application layer is not integrated. The bandwidth allocation algorithm is combined with flow control in IP networks. In fact, P2P applications have become the largest consumer of IP network resources, largely exceeding the data traffic of Web, E-mail, FTP, etc. and becoming the main burden on the backbone network, even causing network congestion, thereby reducing and affecting performance of other services. This patent combines fair bandwidth allocation of peer-to-peer network with bandwidth allocation of IP network link, realizes cross-layer bandwidth allocation of P2P network and IP network, and effectively controls backbone network while satisfying node resource sharing and downloading in peer-to-peer network. network traffic, avoid backbone network congestion, and improve network performance.

SUMMARY OF THE INVENTION

The purpose of the invention is to provide a cross-layer P2P resource sharing network bandwidth allocation algorithm with fair and reasonable allocation, effective sharing and downloading of node resources, and control of backbone network traffic at the same time.

In order to achieve the above object, the following technical solutions are adopted: the algorithm of the present invention mainly includes a P2P peer-to-peer network, a service requester s, a service provider p and an IP network link 1, and in the P2P peer-to-peer network, the service requester s needs to obtain For the services provided by the service provider p, such as file sharing and downloading, the service provider p must allocate its upload bandwidth to the service requester s, and complete the data transmission service through the existing IP network. The service provider p calculates the optimal bandwidth allocated by the service provider p to the service requester s according to the price paid by the service requester s and the price charged by the link l.

The algorithm steps are as follows:

Step 1, in the P2P peer-to-peer network, the service provider p that provides the download service initializes the bandwidth allocation, and initializes the allocated bandwidth x _ps [t] for each service requester s at time t;

Step 2, if the bandwidth allocation x _ps [t] is already the optimal point of the resource allocation model at this time, the algorithm ends, and the service provider p transmits data to the service requester s at this rate, otherwise, go to step 3;

Step 3, at time t, each service requester s calculates the aggregated bandwidth y _s [t] obtained by it according to the bandwidth x _ps [t] allocated to it by each service provider p;

In the formula, s is the service requester; p is the service provider; P(s) is the set of service providers that provide download services for the service requester s;

Step 4: At time t, each service requester s calculates the price λ _s [t] that it should pay to the entire network according to the aggregated bandwidth y _s [t] obtained by it;

In the formula, ws is the fee paid by the service requester _s ; ε > 0 is a small positive number, so as to avoid the price λ _s [t] paid by the service requester s from being too large;

Step 5, at time t, the link l in the IP network initializes the price μ _l [t] it charges;

Step 6, the link l in the IP network calculates the aggregated traffic z _l [t] on the link according to the traffic x _ps [t] passing through its link;

是0-1函数，若服务提供者p为服务请求者s提供服务时的流量经过链路l，则

为1，否则

为0；where S(p) is the set of all service requesters that service provider p provides download services; P is the set of all service providers; x _ps [t] is the bandwidth allocated by service provider p to service requester s ,

is a 0-1 function, if the traffic when the service provider p serves the service requester s passes through the link l, then

is 1, otherwise

is 0;

Step 7, link l adjusts the price μ _l [t+1] charged at time t+1 according to the aggregated traffic z _l [t] on the link and its own link bandwidth C _l ;

In the formula, β>0 is the iterative step size; C _l is the bandwidth of link l;

意味着，

mean,

If μ _l [t]>0, then

If μ _l [t]=0, then

Step 8: According to the price λ _s [t] paid by the service requester to the network and the price μ _l [t] charged by the link l in the IP network, the actual price γ _ps [ t];

之差；where L(p,s) is the path from the service provider p to the service requester s, which is a non-empty subset of the link set L; the price γ _ps [t] is what the service requester s pays to the network The price _λs [t] and the price charged for the transmission path from the service provider p to the service requester s

Difference;

Step 9, the service provider p calculates the service provider according to the bandwidth x _ps [t] allocated to each service requester s and the price γ _ps [t] paid by each service requester s to the service provider p the expected price of p ξ _p [t];

In the formula, S(p) is the set of all service requesters that the service provider p provides download services; x _ps [t] is the bandwidth allocated by the service provider p to the service requester s; γ _ps [t] is the service request is the actual price paid by s to service provider p; C _p is the upload bandwidth of service provider p;

Step 10, the service provider p according to the bandwidth x _ps [t] allocated to each service requester s, and the price γ _ps [t] paid by each service requester s to the service provider p, and the service provider p The expected price ξ _p [t] of , adjust the bandwidth x _ps [t+1] allocated to the service requester s at time t+1

是对时刻t带宽分配x_ps[t]的估值；

是对服务提供者p为服务请求者s分配的带宽x_ps[t+1]的估值；θ是低通滤波因子，且0＜θ＜1，能够有效消除最优点不唯一而引起的算法波动问题；In the formula, α>0 is the iteration step size; x _ps [t] is the bandwidth allocated by the service provider p to the service requester s;

is an estimate of the bandwidth allocation x _ps [t] at time t;

is the estimation of the bandwidth x _ps [t+1] allocated by the service provider p to the service requester s; θ is the low-pass filter factor, and 0<θ<1, which can effectively eliminate the algorithm caused by the non-unique optimal point volatility problem;

意味着

mean

If x _ps [t] > 0, then

If _xps [t]=0, then

Step 11, each service provider p iterates according to steps 3 to 10 until it reaches the optimal point, that is, the optimal allocation of the upload bandwidth of the service provider among the service requesters;

Step 12, if a new service provider or service requester joins in the network, or an original service provider or service requester quits, the iterative process steps 1 to 10 are repeated until a new optimal point is reached.

Compared with the prior art, the present invention has the following advantages:

1. The algorithm of the present invention can effectively converge to the optimal point of upload bandwidth allocation of service providers in the P2P resource sharing network, so as to realize fair allocation of the upload bandwidth of service providers among service requesters.

2. Considering the integration of the P2P network and the IP backbone network, the optimal bandwidth allocation based on the idea of cross-layer optimization is realized. While satisfying the node resource sharing and downloading in the peer-to-peer network, it can effectively control the traffic in the backbone network and avoid the backbone network. Network congestion, improve network performance.

Description of drawings

FIG. 1 is a flowchart of the P2P client of the algorithm of the present invention.

FIG. 2 is a flow chart of the IP link end of the algorithm of the present invention.

Fig. 3 is a topological structure diagram of 6 user networks of the algorithm of the present invention.

FIG. 4 is an optimal bandwidth allocation diagram obtained by the service requester s1 in FIG. 2 .

FIG. 5 is an optimal bandwidth allocation diagram obtained by the service requester s2 in FIG. 2 .

FIG. 6 is an optimal bandwidth allocation diagram obtained by the service requester s3 in FIG. 2 .

FIG. 7 is an optimal bandwidth allocation diagram obtained by the service requester s4 in FIG. 2 .

Figure 8 is a graph of network aggregation utility at different network sizes.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings:

The algorithm of the present invention mainly includes a P2P peer-to-peer network, a service requester s, a service provider p and an IP network link l. In the P2P peer-to-peer network, the service requester s wants to obtain the services provided by the service provider p, such as file sharing. For this purpose, the service provider p allocates its upload bandwidth to the service requester s, and completes the data transmission service through the existing IP network. The service provider p calculates the optimal bandwidth allocated by the service provider p to the service requester s according to the price paid by the service requester s and the price charged by the link l.

The algorithm steps are as follows:

Step 1, in the P2P peer-to-peer network, the service provider p that provides the download service initializes the bandwidth allocation, and initializes the allocated bandwidth x _ps [t] for each service requester s at time t;

Step 2, if the bandwidth allocation x _ps [t] is already the optimal point of the resource allocation model at this time, the algorithm ends, and the service provider p transmits data to the service requester s at this rate, otherwise, go to step 3;

Step 3, at time t, each service requester s calculates the aggregated bandwidth y _s [t] obtained by it according to the bandwidth x _ps [t] allocated to it by each service provider p;

In the formula, s is the service requester; p is the service provider; P(s) is the set of service providers that provide download services for the service requester s;

Step 4: At time t, each service requester s calculates the price λ _s [t] that it should pay to the entire network according to the aggregated bandwidth y _s [t] obtained by it;

In the formula, ws is the fee paid by the service requester _s ; ε > 0 is a small positive number, so as to avoid the price λ _s [t] paid by the service requester s from being too large;

Step 5, at time t, the link l in the IP network initializes the price μ _l [t] it charges;

Step 6, the link l in the IP network calculates the aggregated traffic z _l [t] on the link according to the traffic x _ps [t] passing through its link;

是0-1函数，若服务提供者p为服务请求者s提供服务时的流量经过链路l，则

为1，否则

为0；where S(p) is the set of all service requesters that service provider p provides download services; P is the set of all service providers; x _ps [t] is the bandwidth allocated by service provider p to service requester s ,

is a 0-1 function, if the traffic when the service provider p serves the service requester s passes through the link l, then

is 1, otherwise

is 0;

Step 7, link l adjusts the price μ _l [t+1] charged at time t+1 according to the aggregated traffic z _l [t] on the link and its own link bandwidth C _l ;

In the formula, β>0 is the iterative step size; C _l is the bandwidth of link l;

意味着，

mean,

If μ _l [t]>0, then

If μ _l [t]=0, then

Step 8: According to the price λ _s [t] paid by the service requester to the network and the price μ _l [t] charged by the link l in the IP network, the actual price γ _ps [ t];

之差；where L(p,s) is the path from the service provider p to the service requester s, which is a non-empty subset of the link set L; the price γ _ps [t] is what the service requester s pays to the network The price _λs [t] and the price charged for the transmission path from the service provider p to the service requester s

Difference;

Step 9, the service provider p calculates the service provider according to the bandwidth x _ps [t] allocated to each service requester s and the price γ _ps [t] paid by each service requester s to the service provider p the expected price of p ξ _p [t];

In the formula, S(p) is the set of all service requesters that the service provider p provides download services; x _ps [t] is the bandwidth allocated by the service provider p to the service requester s; γ _ps [t] is the service request is the actual price paid by s to service provider p; C _p is the upload bandwidth of service provider p;

Step 10, the service provider p according to the bandwidth x _ps [t] allocated to each service requester s, and the price γ _ps [t] paid by each service requester s to the service provider p, and the service provider p The expected price ξ _p [t] of , adjust the bandwidth x _ps [t+1] allocated to the service requester s at time t+1

是对时刻t带宽分配x_ps[t]的估值；

是对服务提供者p为服务请求者s分配的带宽x_ps[t+1]的估值；θ是低通滤波因子，且0＜θ＜1，能够有效消除最优点不唯一而引起的算法波动问题；In the formula, α>0 is the iteration step size; x _ps [t] is the bandwidth allocated by the service provider p to the service requester s;

is an estimate of the bandwidth allocation x _ps [t] at time t;

is the estimation of the bandwidth x _ps [t+1] allocated by the service provider p to the service requester s; θ is the low-pass filter factor, and 0<θ<1, which can effectively eliminate the algorithm caused by the non-unique optimal point volatility problem;

意味着

mean

If x _ps [t] > 0, then

If _xps [t]=0, then

Step 11, each service provider p iterates according to steps 3 to 10 until it reaches the optimal point, that is, the optimal allocation of the upload bandwidth of the service provider among the service requesters;

Step 12, if a new service provider or service requester joins in the network, or an original service provider or service requester quits, the iterative process steps 1 to 10 are repeated until a new optimal point is reached.

The invention combines the bandwidth allocation algorithm of the application layer with the flow control in the IP network. In fact, P2P applications have become the largest consumer of IP network resources, largely exceeding the data traffic of Web, E-mail, FTP, etc. and becoming the main burden on the backbone network, even causing network congestion, thereby reducing and affecting performance of other services. This patent combines the fair bandwidth allocation of the peer-to-peer network with the bandwidth allocation of the IP network link to realize a network bandwidth fair allocation scheme based on cross-layer optimization. While satisfying the node resource sharing and downloading in the peer-to-peer network, it can effectively control the backbone network traffic, avoid backbone network congestion, and improve network performance.

The cross-layer P2P resource allocation model is actually a cross-layer optimization problem, and the flow control of the application layer and the congestion control of the transport layer are studied at the same time. Three different entities are included here, namely the service requester (ie the client), the service provider (ie the server) and the service bearer (ie the link).

The cross-layer P2P resource allocation model is:

是0-1函数，若服务提供者p为服务请求者s提供服务时的流量经过链路l，则

为1，否则

为0；U_s(y_s)是服务请求者s的效用函数，代表服务请求者s获得下载带宽y_s后的满意度，为了实现服务提供者的上传带宽在服务请求者之间的公平分配，这里选择U_s(y_s)＝w_s log y_s，其中w_r描述了服务请求者s愿意支付的费用，该效用函数实现了服务提供者的上传带宽在服务请求者之间分配的比例公平性。In the formula, s is the service requester (ie the client); p is the service provider (ie the server); l is the service bearer (ie the link) in the network; S is the set of service requesters; P is the service provider L is the set of links in the network; P(s) is the set of service providers that provide download services for service requester s; S(p) is the set of all service requesters who provide download services for service provider p Set; x _ps is the bandwidth obtained by service requester s from service provider p, then the total bandwidth obtained by service requester s is y _s , and at the same time, the sum of bandwidth provided by service provider p for all service requesters does not exceed its own upload bandwidth C _p ; and C _l is the bandwidth of link l,

is a 0-1 function, if the traffic when the service provider p serves the service requester s passes through the link l, then

is 1, otherwise

is 0; U _s (y _s ) is the utility function of the service requester s, representing the satisfaction of the service requester s after obtaining the download bandwidth y _s , in order to realize the fair distribution of the upload bandwidth of the service provider among the service requesters , here we choose U _s (y _s )= _{ws log y s ,} where _wr describes the fee that the service requester s is willing to pay, and the utility function realizes the proportion of the upload bandwidth of the service provider distributed among the service requesters fairness.

l∈L(p,s)，否则

In the multi-source download P2P business, a service requester can download the same file from multiple service providers, thus forming multiple data streams destined to the same service requester, and each data stream has its own transmission path. To this end, define the path from service provider p to service requester s as L(p,s), which is a non-empty subset of link set L. If link l is on path L(p,s), but

l∈L(p,s), otherwise

而对于变量x_ps则不是严格凸优化的，每个服务请求者s会存在多个不同的最优带宽分配

Here, maximizing the aggregated utility of all service requesters is the goal of the entire P2P network. This problem has constraints on the upload bandwidth of the service provider and the transmission bandwidth of the network link. For the above utility function U _s (y _s ) = w _s log y _s , then the resource allocation problem is a strictly convex optimization problem for the variable y _s , and each service requester s has a globally optimal aggregated bandwidth value

For the variable x _ps , it is not strictly convex optimization, and there will be multiple different optimal bandwidth allocations for each service requester s

The Lagrangian function of the above resource allocation problem is

where λ _s , ν _p and μ _l are Lagrangian factors, and n _p and _ml are slack variables. Here, the Lagrangian factor can be interpreted as the price per unit bandwidth in the network, then λ _s is the unit bandwidth price paid by the service requester s, ν _p is the unit bandwidth price charged by the service provider p, and μ _l is the chain The price per unit bandwidth charged by the channel.

Three sub-problems can be obtained by decomposing the above Lagrangian function:

CLIENT(s)

max U _s (y _s )-λ _s y _s

over y _s ≥0, s∈S

SERVER(p)

over x _ps ≥0,s∈S,p∈P

LINK(l)

max μ _l (C _l -m _l )

over μ _l ≥0,l∈L

The practical implications of the above three sub-problems can be given from an economic point of view:

Sub-problem CLIENT(s): Since each service requester in the P2P network is selfish, it wants to maximize its utility, and the utility depends on the aggregate bandwidth y _s it obtains. At the same time, when the service requester obtains the corresponding bandwidth, it has to pay the fee for using the bandwidth. Since λ _s can be understood as the fee per unit bandwidth paid by the user, then U _s (y _s )-λ _s y _s is the benefit obtained by the service requester s, that is, the difference between the utility obtained by the user and the fee paid .

Subproblem SERVER(p): λ _s x _ps is the fee paid by the service requester s when the bandwidth provided by the service provider p is x _ps . ∑ _l:l∈L(p,s) μ _l is the price charged for the path L(p,s) from service provider p to service requester s, x _ps ∑ _l:l∈L(p,s) μ _l is the fee charged by the path for the service requester s to carry the transport service. The actual meaning of this sub-problem is to maximize the benefits obtained by providing file download services for service requesters under the premise of a certain upload bandwidth.

Subproblem LINK(l): The relaxation factor m _l is the remaining bandwidth of link l, then C _l -m _l is the bandwidth that has been allocated to each service requester s for data transmission. Since μ _l is the price per unit bandwidth charged by link l, μ _l (C _l −m _l ) is the revenue of link l.

Convergence Analysis

Convergence is an important indicator to measure the performance of an algorithm. The present invention first considers the situation of two service providers p1, p2 and four service requesters s1, s2, s3, and s4. As shown in FIG. 3, the upload bandwidths of the service providers p1 and p2 are The link bandwidth in the IP backbone network is 1000Mbps, and the payment fees for service requesters s1, s2, s3, and s4 are 5, 10, 15, and 20, respectively. The algorithm step size is α=β=0.05, the filter factor θ=0.2, and the parameter ε =0.01.

The present invention analyzes the fair distribution of the upload bandwidth of the service provider among the service requesters. The simulation results are shown in Figures 4, 5, 6, and 7. For example, in Figure 7, when the number of iterations is only 50, The algorithm has converged to the optimal point. At this time, the optimal bandwidth allocated by the service provider p1 to the service requester s4 is 8.1704Mbps, and the optimal bandwidth allocated by the service provider p2 to the service requester s4 is 9.8296Mbps, while the service requester The optimal total bandwidth obtained by s4 is 18Mbps. Therefore, the algorithm can effectively converge to the optimal point of the resource allocation model within a limited number of iterations.

Next, analyze the performance of the algorithm when the number of nodes in the network increases. At this time, the upload bandwidth of the service provider is 20Mbps, and the payment fee provided by the service requester is 10. The simulation results are shown in Figure 8. It can be found that as the network scale (the number of service requesters and service providers in the network) increases, the algorithm can still effectively converge to the optimal point of the resource allocation model.

The above-mentioned embodiments are only to describe the preferred embodiments of the present invention, and do not limit the scope of the present invention. On the premise of not departing from the design spirit of the present invention, those of ordinary skill in the art can Such deformations and improvements shall fall within the protection scope determined by the claims of the present invention.

Claims (1)

1. A cross-layer P2P resource sharing network bandwidth fair allocation algorithm mainly comprises a P2P peer-to-peer network, a service requester s, a service provider P and an IP network link l, and is characterized in that in the P2P peer-to-peer network, the service requester s needs to obtain the service provided by the service provider P, the service provider P needs to allocate the uploading bandwidth for the service requester s, and the data transmission service is completed through the existing IP network; the service provider p calculates the optimal bandwidth distributed by the service provider p for the service requester s according to the price paid by the service requester s and the price charged by the link l;

the algorithm comprises the following steps:

step 1, in P2P peer-to-peer network, service provider P providing download service initializes bandwidth allocation, initializes allocated bandwidth x for each service requester s at time t_ps[t]；

Step 2, if the bandwidth allocation x is carried out at this time_ps[t]If the optimal point of the resource allocation model is reached, the algorithm is ended, the service provider p transmits data to the service requester s according to the rate, otherwise, the step 3 is carried out downwards;

step 3, at the time t, each service requester s distributes the bandwidth x to each service provider p according to the service providers_ps[t]Calculating the resultant aggregate bandwidth y_s[t]；

Wherein s is a service requestor; p is a service provider; p(s) is a set of service providers that provide download services for service requesters s;

step 4, at the time t, each service requester s obtains the aggregate bandwidth y according to the aggregate bandwidth_s[t]Calculating the price lambda that it should pay to the whole network_s[t]；

In the formula, w_sIs the cost paid by service requester s; > 0 is a small positive number, thereby avoiding the price λ paid by the service requester s_s[t]Too large;

step 5, at time t, link l in IP network initializes its price μ_l[t]；

Step 6, the link l in the IP network is according to the flow x passing through the link_ps[t]Calculating to obtain the aggregation flow z on the link_l[t]；

Wherein, s (p) is the set of all service requesters of the service provider p providing the download service; p is the set of all service providers; x is the number of_ps[t]Is the bandwidth allocated by the service provider p for the service requestor s,

is a function of 0-1, if the traffic of the service provider p serving the service requester s passes through the link l

Is 1, otherwise

Is 0;

step 7, the link l is according to the aggregation flow z on the link_l[t]Link bandwidth C with itself_lAdjusting the price μ charged at time t +1_l[t+1]；

Where β > 0 is the iteration step size C_lIs of link lA bandwidth;

it is meant that,

mu.s of_l[t]If greater than 0, then

Mu.s of_l[t]When the value is equal to 0, then

Step 8, the service requester s pays the price lambda of the network according to it_s[t]And price mu charged by link l in IP network_l[t]Calculating the actual price gamma to be paid to the service provider p_ps[t]；

Where L (p, s) is the path from service provider p to service requestor s, which is a non-empty subset of link set L, and the price γ_ps[t]Is the price lambda paid by the service requester s to the network_s[t]Price charged to transmission path from service provider p to service requester s

The difference between the two;

step 9, the service provider p allocates a bandwidth x to each service requester s_ps[t]And a price gamma paid by each service requester s to service provider p_ps[t]The expected price ξ of the service provider p is calculated_p[t]；

Wherein S (p) is all service requests of the service provider p for providing download servicesA set of requesters; x is the number of_ps[t]Is the bandwidth allocated by the service provider p for the service requestor s; gamma ray_ps[t]Is the actual price paid by the service requester s to the service provider p; c_pIs the upload bandwidth of service provider p;

step 10, the service provider p allocates a bandwidth x to each service requester s_ps[t]And a price gamma paid by each service requester s to service provider p_ps[t]And expected price ξ of service provider p_p[t]Adjusting the bandwidth x allocated to the service requester s at time t +1_ps[t+1]

Where α > 0 is the iteration step size, x_ps[t]Is the bandwidth allocated by the service provider p for the service requestor s;

is to allocate x bandwidth to time t_ps[t]An estimate of (2);

is the bandwidth x allocated to the service provider p for the service requestor s_ps[t+1]An estimate of (2); theta is a low-pass filtering factor, and theta is more than 0 and less than 1, so that the problem of algorithm fluctuation caused by non-unique optimal points can be effectively solved;

means that

If x_ps[t]If greater than 0, then

If x_ps[t]＝0，Then

Step 11, each service provider p iterates according to the steps 3 to 10 until an optimal point is reached, namely the service provider uploads the optimal allocation of bandwidth among the service requesters;

and step 12, if a new service provider or a service requester is added in the network or the original service provider or the original service requester is quitted, repeating the steps 1 to 10 of the iterative process until a new optimal point is reached.

Images