负载控制的网格资源调度  

Grid resource scheduling based on load control

作　　者：毕可军[1,2] 段富海 马满福[1,2] 

机构地区：[1]西北工业大学自动化学院,西安710072 [2]兰州飞行控制有限责任公司,兰州730070

出　　处：《计算机应用》2009年第10期2635-2637,2651,共4页journal of Computer Applications

基　　金：教育部科学技术研究重点项目(208148);甘肃省科技攻关项目(2GS064-A52-035-03)

摘　　要：针对网格环境下的系统负载管理,将网格系统模型描述为节点和通信链路组成的无向图。在此基础上,提出了节点和通信负载的度量方法,给出了相应的阈值定义以及网格负载水平的计算方法。由此构成了资源选择中对节点、通信链路的选择条件,实现节点、通信的负载控制;网格负载水平则是判定拒绝服务的条件,控制系统任务总量。针对该方法,设计了实现管理的体系结构,讨论了模型计算的时间周期,给出了实现控制的调度算法。实验结果表明,该方法在提高系统吞吐量、控制调度失败率等方面表现出良好的性能,特别是在重载情况下,提高了系统的运行效率,增强了系统的健壮性。Concerning the load system management of grid, this paper described grid mode/ as an undirected graph （ i- DG） by computing nodes connected by a set of colnmunicating edges. Therefore, the authors presented the load measurement methods of node and communication, and brought tbrward how to get the load threshold and grid system load grade. By the load ttlreshold, scheduler selected the resource whose node load was under the threshold and that communication load allowed; at the same time, the grid system load grade decided a resource request to be refused or accepted. As the requirement, the authors proposed the load management architecture, analyzed refresh periods of load threshold and grid system load grade. At last, the load management scheduling algorithm was described based on the load threshold and grid system load grade. Finally simulations were pertormed to compare the performance of load management scheduling algorithm with that of the None Minimisation. The experiments show that the load management scheduling algorithm is efficient and robust in controlling failed scheduling ratio and improving the throughput within grid environments, thus it is especially fit for overload.

关 键 词：网格 负载控制 负载阈值 控制算法 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]