节点文献

网格环境下效用启发式任务调度算法研究

Research on Utility Heuristic Task Scheduling Algorithm in Grid Environment

【作者】 高峰

【导师】 万晓冬;

【作者基本信息】 南京航空航天大学 , 测试计量技术及仪器, 2009, 硕士

【摘要】 作为网格计算的一个重要组成部分,任务调度算法的好坏直接影响到网格计算的性能优劣。随着网格应用逐步走向商业化,网格任务调度系统面临着全新的、更高的要求。面向商业应用的网格任务调度系统要求既能最大程度地满足用户需求,获得较高的用户满意度,又能保证网格系统的高吞吐率和稳定性。基于服务质量(QoS)的任务调度算法成为该领域的一个重要研究方向。本文首先研究了网格技术、网格任务调度相关概念以及网格QoS层次结构模型,在此基础上提出了用户效用的概念。然后针对经典动态任务调度算法的不足,并借鉴MCT和Sufferage算法的调度思想,提出了用户效用启发式网格元任务调度算法:MUU在线调度算法和User USufferage批调度算法。这两种算法只考虑满足用户需求,会严重影响网格系统吞吐率和负载均衡性能。为了改善它们的不足,本文引入负载QoS和全局效用的概念,对MUU算法和User USufferage算法加以改进,从而引出了以全局效用为启发式的网格元任务调度算法:MGU在线调度算法和Global USufferage批调度算法。本文研究了主流的网格调度算法仿真工具GridSim和SimGrid,针对两工具的不足,设计并实现了轻量级的网格调度模拟器GridSche。使用该模拟器对本文提出的几种算法的性能在各种情况下进行了全面的仿真比较。实验表明,通过调整负载权重参数,MGU算法和Global USufferage算法既能在一定程度上尽量满足用户需求,又保证网格系统在较高的吞吐率状态下稳定工作。

【Abstract】 As an important part of the grid computing, task scheduling arithmetic has a direct impact on its performance. And with the commercialization of grid computing, task scheduling system is faced with new and higher demands. A commercialized grid task scheduling system requires to maximize to meet the user demands, to win higher user satisfaction, and meanwhile ensure the system throughput and stability. For this reason, the QoS-oriented task scheduling algorithm becomes an important research direction of this field.In this text, grid technology, grid task scheduling concept and grid QoS layering model are presented firstly, based on which, the concept of User Utility are put forward. Then, the User Utility heuristic scheduling arithmetics for grid meta-task, MUU online scheduling arithmetic User USufferage batch scheduling arithmetic, are invented, considering the defects of classic dynamic scheduling arithmetics, and borrowing ideas from MCT and Sufferage scheduling arithmetics. The two arithmetics consider only to satisfy user requirements, and would lead to worsen of system throughput and stability. Considering the defects of them, we bring forward concept of load QoS and Global Utility. After that,based on MUU and User Utility arithmetics, the Global Utility heuristic scheduling arithmetics for grid meta-task, MGU online scheduling arithmetic and Global USufferage batch scheduling arithmetic, are invented.This text researches the popular grid scheduling arithmetic simulation tools, GridSim and SimGrid. And a new lightweight grid scheduling simulation tool is programmed, focused on the defects of the two simulation tools mentioned above. After that, we use this new tool to make comprehensive contrast experiments for the scheduling arithmetics mentioned in this article. The result of these experiments indicate that, MGU and Global USufferage scheduling arithmetics can meet the needs of users to some extent, and meanwhile ensure the system throughput and stability of grid system, by adjusting the load weight parameter.

【关键词】 网格元任务调度服务质量全局效用
【Key words】 gridmeta-taskschedulingQoSGlobal Utility
节点文献中: 

本文链接的文献网络图示:

本文的引文网络