一种快速网格任务调度策略  

A Fast Algorithm for Grid Job Scheduling

作　　者：王莉[1] 窦旻 刘宗田[1] 黄美丽[1] 

机构地区：[1]上海大学计算机工程与科学学院,上海200072 [2]高能物理研究所计算中心网络安全组,北京100049

出　　处：《计算机科学》2007年第6期128-130,共3页Computer Science

基　　金：国家自然科学基金(60575035)资助。

摘　　要：网格任务调度目标有很多,如用户要求任务轮转时间短、花费代价小,而资源提供者希望资源利用率高等,这些目标相互冲突,因此网格任务调度不仅是一个NP难问题,而且是一个多目标优化问题。本文根据网格环境下任务的时间相关性特点,对传统蚁群算法进行了改进,提出了一种快速网格任务调度算法。该算法不仅解决了网格调度中多目标优化问题,而且依据任务调度历史信息生成蚁群算法的初始信息素分布,提高了蚁群算法的求解速度。In a grid environment, the goals of the user and the resource provider can be conflicting, that is, the user may have the goals of minimizing turn-around time and the cost for the job, while a resource provider may have the goals of improving the throughput of the resources. So grid job scheduling is not only an NP-Hard problem, but also a multi-object optimization problem. In this paper, we propose a grid scheduler based on Ant Colony Algorithm, which satisfies the mutually conflicting goals of the users and the resource provider, as well as improves the velocity of the scheduler, by taking advantage of historical information on scheduling to form the initial information pheromone.

关 键 词：网格计算 任务调度 蚁群算法 多目标优化 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]