【摘要】 大规模和超大规模计算系统中的通讯延迟成为影响负载均衡性能的一个重要因素,且延迟具有时变性,而传统的负载均衡策略通常假设通讯固定不变或不考虑通讯延迟开销.本文考虑了系统的通讯延迟开销和延迟时变性特征,给出一种基于广义神经网络(GNN)的层次结构负载均衡策略.该策略具有三个特点:(1)通讯优化的层次结构能够降低大规模计算系统的负载均衡开销;(2)考虑了节点计算速率及通讯延迟时变特性;(3)通过延迟预测可优化任务的通讯延迟和迁移延迟开销.仿真实验验证了该策略在通讯和负载均衡开销方面的性能.更多还原

【Abstract】 Traditional load balancing strategies generally assume that communication delay is deterministic or the communication overhead can be ignored.A hierarchical load balancing strategy based on generalized neural network(GNN),considering the communication overhead and time-varying delay feature,is presented for large-scale computing systems.The strategy possesses three features:1)load balancing overhead can be reduced with optimizing hierarchical structure communication;2)considering the heterogeneity in the processing rates of the nodes and the delay randomness imposed by the communication medium;3)optimizing task communication delay and migration delay.Simulation results demonstrate the capabilities.更多还原