分布式视频信息存储网络的任务均衡分配

Task Distribution Balancing for Distributed Storage Networks of Video Information

考虑将广义超级市场模型(Generalized Supermarket Model,GSM)应用于分布式视频信息存储网络的任务均衡分配。设:①分布式视频信息存储网络由在地理上分布的存储服务结点、存储负载均衡调度结点和客户结点组成;②视频信息存储任务的到达是一泊松过程,任务粒度为视频流文件;③存储任务的执行时间服从指数分布;④系统在初始时刻的存储任务有限;⑤与存储任务的执行时间相比,存储任务的调度时间不计。在概率意义下,深入讨论了基于广义超级市场模型的负载均衡调度与随机服务选择两种情况下的平均存储任务消耗时间,分析结果表明:在大规模分布式视频信息存储网络中,采用广义超级市场模型的存储任务均衡调度可显著提高系统的平均存储任务消耗时间性能。

The Generalized Supermarket Model(GSM) is used to the task distribution balancing for distributed stor- age networks of video information. Suppose that: ①a storage network of video information consists of certain dis- tributed storage service nodes,a storage load balance-scheduling node, and client nodes;②the arrivals of video infor- mation storage tasks can be regarded as a Poisson process,and the storage task granularity is of video stream file;③ the executing time for a storage task is exponentially distributed;④at the initial time,the number of storage tasks is finite;⑤compared with the executing time used for storage tasks,the scheduling time used for storage tasks is ne- glected. In the sense of probability,the mean time consumption of storage tasks in two cases,using Generalized Super- market Model based scheduling strategy and random service node choice based scheduling strategy,is analyzed in de- tail. The result shows that: for large-scale distributed storage networks of video information, the performance of mean time consumption for storage tasks can be increased evidently by using of the GSM-based load balance schedul- ing strategy.