一种自适应动态负载均衡算法

An Adaptive Dynamic Load Balancing Algorithm

负载均衡问题是一个经典的组合优化难题,该文建立了一个集群中的负载均衡问题模型,并提出了一种旨在最小化负载均衡开销的动态自适应算法。由于集群中存在网络延时,所以负载重分配的开销很大一部分取决于CPU间发送、接受的消息的最大数量。该负载均衡算法以最小化负载重分配时CPU间消息发送、接受的数量为目标,根据过载、轻载CPU数量的变化动态调用D算法和R算法,以降低负载均衡开销。

Load balancing is a classical problem of combination,in this thesis,the authors propose a model of load balancing and propose an adaptive dynamic algorithm for minimizing the load redistribute cost in the cluster.Due to the importance of the network latency in the cluster,the redistribution cost is primarily depending on the maximum number of messages sent or received by a processor.The algorithm aims to minimizing maximum number of messages sent or received by a processor.The algorithm call D algorithm or R algorithm according the difference between overload processor and load-deficiency processor for reducing the cost of load redistribute.