基于多维QoS的集群作业调度算法  

Cluster Job Scheduling Algorithm Based on Multiple Dimension QoS

作　　者：赵姗[1] 周兴社[1] 王云岚[1] 

机构地区：[1]西北工业大学高性能计算研究与发展中心,西安710072

出　　处：《计算机工程》2005年第22期70-72,共3页Computer Engineering

摘　　要：调度技术是集群作业管理系统的关键技术之一。但是,目前大多数集群作业管理系统采用的调度技术,很少按照应用对系统资源的需求加以详细区分和权衡,无法同时满足不同应用在高吞吐量、高可用性和实时性等方面的要求,造成了资源的闲置和资源因为负载过重而无法正常工作。该文正是注意到了这种不足,根据不同应用的需求和实时收集到的资源信息,引入了多维QoS的概念,并给出了多维QoS的数学模型,提出了一种基于多维QoS的集群作业调度算法JSBMQ,充分提高了系统的资源利用率和平均响应时间,使集群可以提供不间断的高效的计算服务。Scheduling technology is the key of cluster job management system. At present, general scheduling used in most of cluster job management systems seldom takes the different resource requirements into consideration, which can not meet the demand of high throughout, high usability and real-time and cause the idle and overload of resources. Based on this deficiency, this paper introduces the concept of QoS, proposes the mode/ of multiple dimension QoS and introduces a job scheduling algorithm based on MQoS（JSBMQ）according to the different application requirements and real-time resource information. This algorithm improves the system utilization and average responding time, makes the cluster provide incessant effective computing services.

关 键 词：集群 QOS 作业调度 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]