Abstract: 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 metascheduler CSF4. The users can take the advantages of the queuing service and cust omized scheduling polices provided by the metascheduler. 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.

Key words: grid computing, metascheduler, WSRF, CSF4, localscheduler