一种基于负载均衡的数据部署算法

A Data Deployment Algorithm Based on Load Balance

针对采用分片存储方式且存储节点异构的媒体分发网络(Media Delivery Network,MDN),提出一种数据部署算法,使分布式节点之间达到负载均衡。算法为每个存储节点定义一个权值来描述其服务能力,通过一种反馈机制,即根据各存储节点偏离平衡状态的程度,来选择要部署数据的节点。实验结果表明:使用算法,不仅在数据增加、删除以及增删混合时,可使系统达到一个较低的负载不均衡度,而且对于各种原因导致的系统负载失衡,经过足够长的时间,都能使系统的负载不均衡度收敛到一个较小的数值。

In this paper, a data deployment algorithm of MDN is given to achieve load balance for the distributed heterogeneous system using striping storage method. The algorithm defines a power value for every peer node to describe its service ability, and then a feedback mechanism is used to select the nodes which deviate most from the equilibrium state to deploy media data. Experiments show that not only in eases of data adding, deleting and both data adding and deleting, system can achieve load balance by using the algorithm, but also for system load imbalance caused by a variety of reasons, the load balance degree of system can be recovered after a sufficient period of time by using this algorithm.