DSF:一种时序约束下的快速数据分发算法  

DSF:A Fast Data Distribution Algorithm under Timing Constraint

作　　者：吴吉庆[1] 彭宇行[1] 刘锋[1] 

机构地区：[1]国防科技大学并行与分布处理国防科技重点实验室,湖南长沙410073

出　　处：《电子学报》2012年第2期365-370,共6页Acta Electronica Sinica

基　　金：国家973重点基础研究发展规划(No.2011CB302601);国家863高技术研究发展计划(No.2009AA01Z142)

摘　　要：时序约束下的大规模数据分发在互联网环境下有着越来越广泛的应用.现有的系统大多采用mesh结构组织结点,并通过运行在请求结点的调度算法来控制数据的分发.但请求结点只依据自身的需求来调度数据,并不能保证系统的整体服务效果.本文以改善整体服务效果为目标,提出一种面向服务结点的调度算法———时序约束下的快速分发算法DSF(Deadline Sensitive Fast distribution).该算法的基本思想是:当服务结点面临多个邻居结点的多个数据请求时,选择系统最迫切需要的数据,优先传输给继续服务能力较强的结点,以达到减少迟到数据比例、提高数据传输率、优化系统持续服务能力的目的.实验结果表明,与面向请求结点的调度算法相比,DSF在流传输质量、分发速率、负载均衡等方面均具有较好的特性.The applications for large-scale data distribution under timing constraint have been widely deployed on Intemet. Most existing systems organize the involved clients with mesh structure and control the data distribution through the scheduling algorithms running on requesting peers. However, requesting peers request data only according to their own needs, regardless of the over- all service quality of the system. Therefore, we take aim at im^oving the overall service quality of the system and propose a scheduling algorithm running on serving peers： DSF （ Deadline Sensitive Fast distribution algorithm）, which transfers urgently needed data to the peers that have strong follow-up service ability with high priority, to reduce the ratio of late data chunks and improve the service capacity of the whole system. The experimental results show that DSF has better performance with respect to transmission quality, distribution rate and load balance than the frequently used scheduling algorithms.

关 键 词：数据分发 时序约束 调度 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]