自主协助动态并行I/O调度策略研究

Research of autonomic assistance dynamic parallel I/O scheduling policy

随着处理器速度和网络传输速度的快速提升,I/O成为限制计算机性能的一大瓶颈,并行I/O提供了解决这一问题的有效途径。在并行I/O的实现中,数据访问调度策略对系统的性能有着重要的影响。文章通过对集群计算中现有I/O数据访问调度策略的研究,针对调度器瓶颈问题,提出了自主协助动态I/O调度策略,实验结果表明了所提出策略的有效性。

With the rapid growth of processor and network transfer speeds, I/O has become a bottleneck limiting computer system＇s performance, while parallel I/O provides an effective way to solve this problem. In realizing parallel I/O, I/O scheduling policy plays a very important role in system＇s high performance.By analyzing existing I/O scheduling algorithms on cluster computing system, a new strategy named the dynamic I/O scheduling algorithm of autonomic assistance is put forward, aiming at scheduler bottleneck. Experimental results demonstrate the efficiency of proposed policy.