一种基于通知机制的计算型网格系统解决方案

本文提出了一套面向高性能计算的网格系统构架解决方案。在上层调度设计方面,利用Globus所提供的通知机制,同步网格系统层调度队列与集群调度队列信息。设计了一种基于蚁群算法的调度策略,实现最优化服务选择。提出了一种“心跳检测”策略,保证了网格系统的可靠性。通过运行15皇后求解问题,分析了“心跳检测”策略对性能的影响,通过网格与集群的测试数据对比与分析,论述其各自的优势和劣势。

A Notification Mechanism Based Solution for Computing Grid System

The paper proposed a solution for the high performance computing grid system. In the design of the high layer scheduling, utilizing the notification mechanism provide by the Globus Toolkit 4.0, synchronizing the information between the scheduling queues in the grid layer and the queues in its underlying cluster. Designed a scheduling strategy based on ant colony algorithm to face the requirements of the dynamic discovering and the alteration in the system, and facilitated the optimizing of the service selection. Fur- thermore, proposed a heart- beat test strategy, to guarantee the reliability of the grid system. At last, through Queen 15 problem solu- tion, analyzed the influence of the heart- beat test strategy to the performance of the system, meaning while, through the test result contrast and analysis, discussed the advantage and shortcoming respectively.