【摘要】 讨论了一种基于对等网络的高性能计算系统的体系结构,提出了一种两级联合调度算法.该算法针对对等网络结构松散、节点稳定性差的缺点,充分考虑负载平衡、志愿机CPU空闲计算能力以及子任务相关要求,分两级进行任务调度,从而分散了调度负载,避免了调度瓶颈,增强了系统的动态扩展性,提高了系统的并行效率.实验表明:随着志愿机个数的增加,系统动态平衡效果增强,加速比增大,进而说明了算法的有效性.更多还原

【Abstract】 The P2HP′s architecture,which is a computing system with high performance based on peer-to-peer(P2P) networks,is discussed and its two-layer united scheduling algorithm(TLUSA) is proposed.Owing to P2P networks′ decentralization and instability,the load-balance,idle computational ability of volunteer and correlative requirement of subtasks was taken into account in TLUSA.The scheduling-loads were decentralized by using this algorithm to avoid scheduling bottleneck,and improves dynamic expansibility and computing efficiency of P2HP.Simulation experiments show that the balance effect and the speedup of the system increase with the rise of volunteers,which verifies the validity of TLUSA.更多还原