A Grid Scheduling Technology For BLAST Calculation Jobs

Grid Computing is a new way for computing in the 21st century, and information biology's main task is to use computer softwares to store, search, and analyze biological information [InfoBio], Biology Research Center in Peking University has developed a net system named WebLab. In order to use grid to compute biology computing, we have developed WebLab Grid Computing Environment.The BLAST of WebLab is concurrent, frequently used and has some big jobs. In some cases, users will not bear the computing time of some jobs in the system. In this thesis, we want to find a good scheduling algorithm and a better way of allocating resources to make the system more powerful, and change these phenomena.Grid scheduling [Grid Sc] deals with Job Scheduling that contains algorithm design,resource expression and resource allocation. In order to find suitable scheduling policy, we must analyze information biology computing and BLAST soft in Grid, and we have to reach such a goal: Make resource allocation better and speed up system throughput rate.The main question is how to use computing resources suitably in Grid Evironment to avoid job overtime. Through the introduction of PBS and GridWay, we can see: Suitable scheduling policy can be made,and just setting up job completion time can avoid resource wasting.In oder to solve the problem, we analyze BLAST software in information biology computing, and raise a Grid computing environment that satisfy the needs.Then we introduce and analyze other services that the grid scheduling service uses. At last,we design a scheduling service and realize a solution for this problem: Monit jobs'states, find jobs with wrong resources, and then move these jobs form wrong resources to better ones, and solve the problem preliminary; to reduce the situation, we use suitable policy to allocate resources for jobs that need resource allocation.Then we design and realize a Grid Scheduling Service based on GT4 using WSRF protocol. We test it through the Grid distributed Environment contains one multi-core PC and IA Cluster, then the result tells us: In the Grid Evironment, using scheduling technology to schedule jobs can use computing resources rationally,and reduce the problem.