一种高效的网格任务管理器组件的设计和实现

A Design of a High-Performance Task Manager Module

随着分布式计算系统的需求不断增长,集群分布式系统在处理作业提交和作业查询时出现了相当大的通讯延时问题,降低了系统性能。Task Broker是基于ACE开发环境开发的一种高效的网格任务管理器组件,它采用改进的Fair-Share算法来计算作业的优先级,通过使用ACE提供的应用接口,实现了以异步方式使用一个或多个线程并发处理多个客户端请求的功能。文中介绍了基于Task Broker分布式并行计算系统的结构;详细阐述了用来计算作业优先级的Fair-Share算法以及Task Broker的模型和实现;并且对Task Broker的性能进行了分析和比较。实验证明,TaskBroker较好地解决了通讯延时问题,大大减轻了系统的性能开销。

The demands on grid computing system are increasing rapidly, causing the problem of considerable communication delay and lowering the system performance. We develop Task Broker, which is a high-performance task manager module, for adaptive communication environment （ACE）. In section 1 of the full paper, we explain the architecture of distributed parallel computing system of Task Broker with the help of Fig. 1. In section 2, we design and implement our Task Broker. Its four subsections are： job submission manager （subsection 2. 1）, task distribution manager （subsection 2.2）, task inquiry manager （subsection 2.3）, and event manager （subsection 2.4）. Finally we give an example in section 3 to evaluate the performance of Task Broker. The experimental results, given in Tables 1 and 2 in the full paper, show preliminarily that communication delay is considerably suppressed and the grid computing system＇s capabilities are considerably raised.