基于Kriging代理模型的动态云任务调度方法  

Dynamic cloud task scheduling method based on Kriging surrogate model

作　　者：李召军[1,2] 李征[2] 王希诚[2] 李克秋[1] 

机构地区：[1]大连理工大学计算机科学与技术学院,辽宁大连116023 [2]大连理工大学工业装备结构分析国家重点实验室,辽宁大连116023

出　　处：《计算机工程与应用》2015年第16期30-35,76,共7页Computer Engineering and Applications

基　　金：国家重点基础研究发展规划(No.2012CB025905);111引智计划(No.B14013);国家自然科学基金(No.11202049);中央高校基本科研业务费专项资金(No.DUT14RC(3)060)

摘　　要：任务调度在云计算环境中发挥着重要作用。提出一种基于Kriging代理模型的动态云任务调度方法。通过对云任务在不同资源组合下的性能表现进行Kriging代理模型建模并优化,从而得到对应于该云任务的最优资源分配方案;利用云平台的API,可动态对该云任务实施资源调度。基于Open Stack开源云平台,对两个工程计算应用进行了任务调度性能测试,结果表明该方法可有效动态调整云任务中的资源配给,按需按优对平台中的云任务进行资源调度。The task scheduling method plays an important role in the various cloud computing environments.A Kriging surrogate model based dynamic task scheduling method in cloud computing is proposed in this paper：with the help of Kriging surrogate model established on the specific resource allocations and the performances accordingly belonging to the cloud tasks,the optimal resource scheduling strategy could be obtained;resources for the cloud tasks will then be reconfigured using the cloud APIs.An Open Stack open source cloud platform with two engineering computing applications is set up for method testing,with the results demonstrating that the proposed method can effectively manage the resource configurations for the cloud tasks dynamically,and the tasks will be scheduled according to their optimal needed resources at the same time.

关 键 词：任务调度 云计算 KRIGING OPEN STACK 

分 类 号：TP393[自动化与计算机技术—计算机应用技术] TU323.3[自动化与计算机技术—计算机科学与技术]