Abstract: For the problem of additional communication overhead in dynamic load balancing process, the interrelated mathematical model based on minimum communication overhead has been built and a new strategy based on greedy algorithm to solve the problem has been proposed in this paper. The new strategy reduces the number of load migration and network traffic in dynamic load balancing process. Simulation results show that the new strategy can obtain better allocation scheme which has smaller communication overhead than greedy strategy.

Key words: genetic algorithm, greedy algorithm, dynamic load balancing, communication overhead

摘要： 针对动态负载均衡过程产生额外通信开销的问题，建立了一种基于最小通信开销的数学模型。在此基础上，提出一种利用遗传算法解决该问题的新策略。该策略可减少负载迁移次数，降低动态负载均衡过程中的网络流量。仿真实验表明，该策略可获得比贪心策略具有更小通信开销的分配方案。

关键词: 遗传算法, 贪心算法, 动态负载均衡, 通信开销