分布式计算平台中混合多应用调度策略的研究

Research of mix multi-application scheduling policies in distributed computing platform

提出与分析了分布式计算平台中几种混合多应用的调度策略,它主要面向多个并行应用之间的调度而不是应用内部的调度,应用内部的调度采用了常见的工作队列容错调度算法。与资源信息有关的调度比较起来,这些调度策略运用到了与资源信息无关的调度方式,这使它们的实现更加简单和容易,更加适合于高挥发性的分布式计算系统。针对各种不同的计算强度、资源可用性、任务粒度来划分实验场景,对各种调度策略进行了评测与比较。实验结果表明,这些调度策略各有优缺点,可以作为评估大规模分布式计算环境下的并行分布式应用的有效策略。

This paper proposed and discussed a set of scheduling algorithms which were able to deal with mix multi-applications in distributed computing platform.It was to deal the problem of scheduling a set of competing parallel applications that were submitted for execution on a distributed computing platform by a community of potentially distinct users.These policies followed a knowledge-free approach that was no information concerning the resources or the applications was available to the scheduler.By means of a large set of operational scenarios obtained by combining four different volunteer computing configurations with various application workloads and intensity,evaluated and discussed the performance of these algorithms.The results show that,although there is no a clear winner among the proposed solutions,all the scheduling policies can be a feasible approach for evaluating parallel applications in large scale distributed computing platform.