流媒体系统中内容分发网络的服务器部署优化

Optimal server placement for content delivery network in streaming system

流媒体系统服务器部署方案需要满足所有小区的流量需求,针对服务器数量及位置的不确定性和不同网络节点之间链路的单位带宽租用费不同的特点,建立基于模拟退火遗传算法的最小费用最大流模型。采用SPFA(shortest path faster algorithm)算法寻找流量分配的最短路径,利用模拟退火遗传算法权衡服务器部署成本和网络租用费,获得最低总费用的服务器部署方案。仿真结果表明,该模型与混合整数规划(MIP)模型相比,小规模下两者结果相近,规模越大,前者优势越趋明显。

The streaming server deployment scheme should meet all district traffic demands.Aiming at the characteristic of the uncertainty of the amount and location of the servers,as well as the differences of bandwidth leasing unit price among the links,a minimum cost maximum flow model based on simulated annealing genetic algorithm was set up.The SPFA(shortest path faster algorithm)flux distribution algorithm was adopted to find the shortest path,and simulated annealing genetic algorithm was used to weigh the cost of the servers and network deployment rental,to gain the server deployment scheme with the lowest total cost.In the simulation experiment,the proposed model was compared with the mixed integer programming.The results are si-milar at small scales,as the scale grows,the advantage of the former one is more obvious.