CN113595801A - Deployment method of edge cloud network server based on task flow and timeliness - Google Patents

Abstract

The invention provides a task flow and timeliness-based edge cloud network server deployment method. Firstly, obtaining communication delay parameters of a wireless access point (AP node) and capacity and cost parameters of all edge servers according to a network scene; then constructing an edge network server deployment cost minimization problem model based on task flow and timeliness, executing an approximate algorithm on all AP nodes in an edge cloud network, solving the edge network server deployment cost minimization problem based on the task flow and timeliness, and calculating deployment decisions of the edge server; and finally, deploying the edge server on part of AP nodes in the edge cloud network by the network operator according to the deployment decision. The method and the device can be suitable for deployment of the edge server in the edge cloud network scene, and the total cost for deploying the edge server is minimized on the premise of meeting the computing resource limit of the edge server and ensuring low communication delay between the edge server and other base stations.

Description

Deployment method of edge cloud network server based on task flow and timeliness

Technical Field

The invention mainly relates to the field of edge computing, in particular to an edge cloud network server deployment method based on task flow and timeliness.

Background

The edge computing is that an open platform integrating network, computing, storage and application is adopted at one side close to an object or a data source, so that services can be provided nearby. In recent years, with the rapid development of the internet of things, the number of network edge terminal devices is increased, the user demand is continuously increased, the data volume generated by the edge devices is rapidly increased, a large amount of sensing data and corresponding real-time processing workload bring important challenges to the practical application of a large-scale internet of things system, and edge computing provides nearby instant storage and data processing services for the nodes of the internet of things with limited resources, so that the edge computing is a solution for effectively coping with the challenges.

With the rapid development of edge computing in the application of the internet of things, more and more terminal devices are added into an edge computing mode, and in order to support the real-time processing of a large-scale internet of things, an edge server with storage and computing capabilities needs to be deployed. The edge server is used as a main calculation carrier of the edge calculation and the edge data center, so that the energy consumption of data calculation can be reduced, the fluency and the rapidness of data transmission can be improved, and the real-time requirements of a client and a server are met. In an edge cloud network, in order to approach a user and reduce communication cost and time delay, an edge server is mainly deployed at the same position as a wireless Access Point (AP) or a base station, when the edge server is deployed, network traffic, deployment cost, computing resource limit of the edge server and the like need to be considered, and the reasonable position for deploying the edge server can be selected to improve network performance and guarantee service stability, so that edge computing service is provided for a terminal device.

In summary, in order to reduce the deployment cost of the edge server in the edge cloud network and meet the requirement of Quality of Service (Quality of Service) of the user application, the edge computing platform is used to deploy the edge server in the edge cloud network, and the deployment cost of the edge server is minimized under the premise of meeting the computing resource limit of the edge server and ensuring low communication delay between the edge server and other base stations.

Disclosure of Invention

The invention provides a task flow and timeliness-based edge cloud network server deployment method which is mainly applied to the aspect of edge computing and has the main advantage that the number of deployed edge servers is minimized under the condition that the maximum communication delay of unit data packets of each wireless network access point (AP node) in an edge cloud network and the limitation of computing resources of the servers are met, so that the total cost for deploying the edge servers is minimized. The scheme of the invention is as follows:

1. the network operator can obtain the deployment scheme of the edge cloud network server through an approximation algorithm:

step 1, constructing an edge cloud network scene, wherein M AP nodes exist in the network, and the AP node set is formed by P ═ P { (P)₁,p₂,...,p_i,...,p_MDenotes that the AP nodes are connected with each other, AP node p_iAnd p_i′T is used as the communication delay between_i,i′Denotes, AP node p_iMaximum communication delay requirement per unit packet to server by d_iIndicating that the cost of deploying servers on any AP node is denoted by cost, p_iUpper maximum traffic c_iMeans that the computing resources of the deployed servers are fixed, denoted by cap ≧ max (c)₁,c₂...,c_i,...,c_M)；

Step 2, for AP node p_iBuilding a coverage set S_i＝{p_i′|t_i,i′≤d_i,p_i′E.g. P, at S_iAt least one AP node needs to be deployed with an edge server;

step 3, constructing a task flow and timeliness-based edge cloud network server deployment problem model, and for all sets S_iThe amount of computing resources of the task admitted at the edge server must not exceed the server computing resourcesThe goal of minimizing the deployment cost of the edge server in the edge cloud network under the constraint of the total amount of the source can be expressed as

Using decision variables x_iIs shown at AP node p_iWhether to deploy a server, x_i1 denotes deployment, x _i0 means no deployment;

step 4, solving the edge cloud network server deployment problem model based on task flow and timeliness in the step 3 by adopting an approximate algorithm to obtain a solution set X ═ X { X } of the deployed edge server₁,x₂,...,x_i,...,x_M}，x_iE {0,1}, and then at x _i1 AP node p_iDeploying an edge server, computing

Resulting in the total cost required to deploy the edge servers.

2. Further, a decision set for deploying servers on the AP nodes is obtained by using an approximate algorithm, which needs to satisfy communication delay between each AP node and an edge server and a total amount limit of server computing resources, and the purpose of minimizing the deployment cost of the edge server is achieved by minimizing the number of deployed edge servers under the condition that the deployment cost of the edge server is the same.

3. Further, the approximation algorithm comprises at least the following steps:

1) for a set of AP nodes P, initialize each AP node P_i Decision variable x _i0, S is { S ═ S₁,...,S_i,...,S_M}，S_iAs AP node p_iCovering the set;

2) for any node p_iE.g. P, calculating the evaluation value

By beta_i＝arctan(μ(p_i) Normalized, count p_iThe number of occurrences in all coverage sets is counted as count (p)_i) Calculating the coverage-to-delay ratio alpha_i＝count(p_i)/β_i；

3) The nodes in the set P are arranged according to alpha_iSequencing and renumbering from big to small;

4) establishing a resource allocation matrix L of | P | × | P | according to the condition that the AP node is connected to the server under the time delay limitation, and aiming at any L [ i |)][j]If t is_i,j≤d_iThen L [ i ]][j]＝c_iOtherwise L [ i ]][j]＝0；

5) Initializing j to 1;

6) for the jth column of the matrix L, let

Δ(j)＝σ_j-cap, if Δ (j) ≦ 0, performing step 8), otherwise performing step 7);

7) for any i ≠ j, let q_i＝|Δ(j)-L[i][j]According to q |_iSelecting the ith row from small to large, and executing the operation that delta (j) is delta (j) -L [ i][j]Update L [ i][j]If delta (j) > 0, continuing to select the next line for judgment, otherwise, executing the step 8);

8) for arbitrary p_iE.g., P, and L [ i][j]If > 0, then execute P ═ P- { P_iH, update all L [ i ] of the ith row][j]Is 0, and x is updated_jIf 1, then

Step 9) is executed, otherwise j equals j +1, and step 6) is executed by jumping;

9) obtaining a decision set X ═ X of server deployment on AP nodes₁,x₂,...,x_i,...,x_MAnd fourthly, finishing the algorithm.

Compared with the prior art, the method has the advantages that:

the deployment method of the edge cloud network server based on the task flow and the timeliness is suitable for terminal equipment groups of different types and different scales, the limit of computing resource capacity of the edge server, the node flow and the communication time delay are considered, the edge server does not need to be deployed on all AP nodes, and the cost for deploying the edge server is reduced as far as possible under the condition that the conditions are met through an algorithm.

Drawings

FIG. 1 is a block flow diagram of the present invention;

FIG. 2 is an algorithmic flow diagram of the edge server deployment method of the present invention;

FIG. 3 is an exemplary diagram of an AP node of the present invention;

FIG. 4 is an exemplary diagram of an edge server deployment location of the present invention;

FIG. 5 is a partial schematic of the algorithm of the present invention;

Detailed Description

The present invention is described in further detail below with reference to fig. 3, 4 and 5.

Assume edge server deployment in public places such as malls and supermarkets in a 5G background as an example.

Step one, obtaining communication delay parameters of all AP nodes in a network and parameters of capacity and deployment cost of all edge servers;

step two, for the AP node p_iBuilding a coverage set S_i＝{p_i′|t_i,i′≤d_i,p_i′E.g. P, at S_iAt least one AP node needs to be deployed with an edge server, and the AP node p_iAnd p_i′T is used as the communication delay between_i,i′Denotes, AP node p_iMaximum communication delay requirement per unit packet to server by d_iRepresents;

step three, determining a target of a network operator for deploying the edge server, namely determining that the deployment cost of the edge server in the edge cloud network is minimized under a constraint condition that the computing resource amount of the task admitted by the edge server in the coverage set does not exceed the total computing resource amount of the server, wherein the target can be expressed as

Using decision variables x_iIs shown at AP node p_iWhether to deploy a server, x_i1 denotes deployment, x_iTable (0)Showing not to deploy, cost is the cost of deploying the edge server;

step four, obtaining a solution set X ═ X of the deployment edge server by adopting an approximate algorithm₁,x₂,...,x_i,...,x_M}，x_iE {0,1}, and then at x _i1 AP node p_iDeploying an edge server, computing

Get the total cost required to deploy the edge servers:

a) for a set of AP nodes P, initialize each AP node P_i Decision variable x _i0, S is { S ═ S₁,...,S_i,...,S_M}，S_iAs AP node p_iThe coverage set, exemplified by fig. 3, assumes that there is a total of 5 AP nodes, with set P ═ P₁,p₂,p₃,p₄,p₅Denotes, node p₁,p₂,p₃,p₄,p₅D for maximum communication delay₁,d₂,d₃,d₄,d₅Represents, node p₁,p₂,...,p₅For communication delay to other nodes (t)_1,2,t_1,3,...,t_1,5),(t_2,1,t_2,3,...,t_2,5),...，(t_5,1,t_5,2,...,t_5,4) Represents;

b) for any node p_iE.g. P, calculating the evaluation value

By beta_i＝arctan(μ(p_i) Normalized, count p_iThe number of occurrences in all coverage sets is counted as count (p)_i) Calculating the coverage-to-delay ratio alpha_i＝count(p_i)/β_i；

c) The nodes in the set P are arranged according to alpha_iSequencing and renumbering in descending order, and taking the example of FIG. 3, sequencing the nodes in the set P to obtain new numbers;

d) establishing a resource allocation matrix L of | P | × | P | according to the condition that the AP node is connected to the server under the time delay limitation, and aiming at any L [ i |)][j]If t is_i,j≤d_iThen L [ i ]][j]＝c_iOtherwise L [ i ]][j]＝0；

e) Initializing j to 1;

f) for the jth column of the matrix L, let

Δ(j)＝σ_j-cap, if Δ (j) is ≦ 0, performing step h), otherwise performing step g);

g) for any i ≠ j, let q_i＝|Δ(j)-L[i][j]According to q |_iSelecting the ith row from small to large, and executing the operation that delta (j) is delta (j) -L [ i][j]Update L [ i][j]If delta (j) > 0, continuing to select the next line for judgment, otherwise, executing the step h);

h) for arbitrary p_iE.g., P, and L [ i][j]If > 0, then execute P ═ P- { P_iH, update all L [ i ] of the ith row][j]Is 0, and x is updated_jIf 1, then

Then step i) is executed, otherwise j is j +1, and the step f) is executed by skipping, as exemplified in fig. 5, after the execution of steps a), b) and c), the set P is updated, and after the execution of steps d), e), f), g) is executed, the nodes P corresponding to the first and second columns of the matrix L are selected after the step h), e)₁,p₂Updating x as a node of a deployment server₁＝1,x₂＝1；

i) Obtaining a decision set X ═ X of server deployment on AP nodes₁,x₂,...,x_i,...,x_MThe algorithm ends, illustrated by fig. 4, the final deployment position of the edge server.

Claims (3)

1. An edge cloud network server deployment method based on task traffic and timeliness is characterized in that a network operator can regard the deployment problem of an edge server as an aggregate coverage problem, obtains a deployment scheme of the edge cloud network server through an approximation algorithm under the condition that the maximum communication delay of a unit data packet of each wireless network access point (AP node) in an edge cloud network and the limitation of server computing resources are met, and minimizes the cost of server deployment, and the method at least comprises the following steps:

step 1, constructing an edge cloud network scene, wherein M AP nodes exist in the network, and the AP node set is formed by P ═ P { (P)₁,p₂,...,p_i,...,p_MDenotes that the AP nodes are connected with each other, AP node p_iAnd p_i′T is used as the communication delay between_i,i′Denotes, AP node p_iMaximum communication delay requirement per unit packet to server by d_iIndicating that the cost of deploying servers on any AP node is denoted by cost, p_iUpper maximum traffic c_iMeans that the computing resources of the deployed servers are fixed, denoted by cap ≧ max (c)₁,c₂…,c_i,…,c_M)；

Step 2, for AP node p_iBuilding a coverage set S_i＝{p_i′|t_i,i′≤d_i,p_i′∈P}；

Step 3, constructing a task flow and timeliness-based edge cloud network server deployment problem model, and for all sets S_iThe objective can be expressed as minimizing the deployment cost of the edge server in the edge cloud network under the constraint that the computing resource amount of the task admitted by the edge server does not exceed the total computing resource amount of the server

Using decision variables x_iIs shown at AP node p_iWhether to deploy a server, x_i1 denotes deployment, x_i0 means no deployment;

step 4, solving the edge cloud network server deployment problem model based on task flow and timeliness in the step 3 by adopting an approximate algorithm to obtain a solution set X ═ X { X } of the deployed edge server₁,x₂,…,x_i,…,x_M}，x_iE {0,1}, and then at x_i1 AP node p_iDeploying an edge server, computing

Resulting in the total cost required to deploy the edge servers.

2. The method for deploying the edge cloud network server based on the task traffic and the timeliness as claimed in claim 1, wherein an approximation algorithm is used to obtain a decision set for deploying the server on the AP node, which is required to meet the communication delay between each AP node and the edge server and the total amount of computing resources of the server, and the purpose of minimizing the deployment cost of the edge server is achieved by minimizing the number of deployed edge servers under the condition that the deployment costs of the edge servers are the same.

3. The method for deploying the edge cloud network server based on the task traffic and the timeliness as claimed in claim 1, wherein the approximation algorithm for solving the server deployment problem at least comprises the following steps:

1) for a set of AP nodes P, initialize each AP node P_iDecision variable x_i0, S is { S ═ S₁,…,S_i,…,S_M}，S_iAs AP node p_iCovering the set;

2) for any node p_iE.g. P, calculating the evaluation value

By beta_i＝arctan(μ(p_i) Normalized, count p_iThe number of occurrences in all coverage sets is counted as count (p)_i) Calculating the coverage-to-delay ratio alpha_i＝count(p_i)/β_i；

3) The nodes in the set P are arranged according to alpha_iSequencing and renumbering from big to small;

4) according to AP node connecting to delay limitationEstablishing a resource allocation matrix L of P x P under the condition of the server, and aiming at any L [ i |)][j]If t is_i,j≤d_iThen L [ i ]][j]＝c_iOtherwise L [ i ]][j]＝0；

5) Initializing j to 1;

6) for the jth column of the matrix L, let

Δ(j)＝σ_j-cap, if Δ (j) ≦ 0, performing step 8), otherwise performing step 7);

7) for any i ≠ j, let q_i＝|Δ(j)-L[i][j]According to q |_iSelecting the ith row from small to large, and executing the operation that delta (j) is delta (j) -L [ i][j]Update L [ i][j]If delta (j) > 0, continuing to select the next line for judgment, otherwise, executing the step 8);

8) for arbitrary p_iE.g., P, and L [ i][j]If > 0, then execute P ═ P- { P_iH, update all L [ i ] of the ith row][j]Is 0, and x is updated_jIf 1, then

Step 9) is executed, otherwise j equals j +1, and step 6) is executed by jumping;

9) obtaining a decision set X ═ X of server deployment on AP nodes₁,x₂,...,x_i,...,x_MAnd fourthly, finishing the algorithm.

Images