一个基于CSF4的多集群网格计算平台的设计与实现

Design and Implementing of a Grid Computing System Based on CSF4

设计并实现了一个基于WSRF的网格计算平台,该系统基于G lobus Toolkit 4,集成了元调度器CSF4,本地调度器SGE5.3/6.0,LSF6.0和openPBS2.3.16以及数据网格系统G farm.在功能上,该系统可协同本地及远程的资源管理器,支持WS-GRAM和Pre-WS-GRAM(GT2 Gatekeeper)的协作,支持完全地委托用户代理证书;在任务管理上,该系统提供了排队服务,并支持定制调度策略;另外,该系统支持LSF,PBS,SGE,Condor等多种资源管理器类型.

The applications in such fields as bioinformatics, etc. increasingly demand more powerful computing environments and huge amounts of distributed data. One single supercomputer or cluster cannot meet the requirements. Using meta-scheduler can coordinate multiple local clusters so that it is able to provide a scalable grid environment with much more computing resources and disk space. We designed a grid infrastructure based on GT4 （WSRF） and a meta-scheduler CSF4. The users can take the advantages of the queuing service and customized scheduling polices provided by the meta-scheduler. CSF4 is enhanced to work with different local schedulers, like LSF, PBS, SGE and Condor etc, in remote clusters. The system supports both GT4 WS-GRAM and GT2 GRAM （Gatekeeper） and is able to delegate client proxy with full delegation. Furthermore, Gfarm is integrated to the system to provide a global file system for data-intensive applications. The system has been deployed on the test-bed of PRAGMA （ Pacific rim applications ＆ grid middleware assembly）. A few of bioinformatic applications, like iGap and wublast, have been run successfully on the system.