Abstract: Load balancing in grid is different from traditional distributed system because of scalability，heterogeneity of computing resources and considerable transfer delay.In this paper，a dynamic and decentralized load balancing algorithm is presented.The processing power and transfer delay are considered in this algorithm.The algorithm is an instantaneous distribution policy that are performance driven to minimize execution cost.The objective of the load balancing algorithm is to minimize the average job response time.The simulation results show that the algorithm deeply reduces the average job response time.

摘要： 网格环境由于其可扩展性、异构性以及大量的传输延迟，使得网格环境下的负载均衡不同于传统的分布式系统。提出了一种动态的分布式负载均衡算法，该算法综合考虑网格站点的处理能力和站点之间的传输延迟，采用即时分配策略来降低作业的执行成本，目标是使系统平均作业响应时间最小化。仿真结果显示该算法显著减少了作业的平均响应时间。