网格计算池模型实现的一些关键技术研究

Research on Some Key Techniques of Computing Pool Model Realization

网格计算池是近年来针对传统“累加”式网格计算模型中网络带宽受限制、通信延迟、通用性不强等缺陷提出的一种新型网格模型。并行计算环境与任务调度管理作为超级计算资源中的重要组成部分,由于其专业性与普及度的影响,长期以来应用的侧重点都更多地在于计算的速度与效率上,在系统开放性上并不注重,对于其二次开发如在网格计算池中的应用更是不多见。就并行计算环境的优化、针对于网格计算池模型应用中的配置和任务调度管理软件以及利用IFL实现网格计算池模型接口中的一些关键问题进行了研究。

In recent years, people hope a grid can accumulate computing power of many supercomputers to form an unprecedented virtual supercomputer,but in fact many imperfections have being seen towards this goal, for example, the shortage of bandwidth, the delay of communication and which only could be used in special applications. The computing pool is a new grid model which is advocated for meliorating these defects. Both parallel computation environment and portable batch system are always the important component of the supercomputer system,but for they are only used for special domain and by special people, there is the more emphasis particularly on computing speed and efficiency than on the opening of the system. The application which based on them such as computing pool also is very few. This paper describes some key technologies used in the computing pool system, such as the optimization of the parallel computation environment, the special configuration for realization of the computing pool system and how to use the portable batch system IFL to realize the interface of the computing pool system.