一种多资源负载平衡算法———RLBA

A multi-resources load-balancing algorithm

讨论了一种适用于分布式系统的多种资源的负载平衡算法。调度时,首先将任务按照其所占用的系统资源比例进行分类,然后在线计算每个节点的综合资源负载,并按照节点与整个系统的平均综合资源负载的差值进行类型划分,最后在负载最小的节点类中按照均匀概率随机选择一个节点作为任务调度的对象。算法还考虑了调度时可能发生的倾泻现象,并给出了避免的方法。

A multi-resources load-balancing algorithm in the distributed system was brought forward. While scheduling, first, the arriving tasks were classified by the resources which they would take up later, and then a load of each node was calculated online. Nodes in the system were also classified into some classes in term of the deference between load of a node and the average value of system. At last, a node in the least loading class was selected randomly and uniformly. The algorithm can avoid the scheduling-pouring effect. It has been adopted in a distributed system and got a satisfying result.