The Research And Implementation On Grid Workflow Scheduling Mechanism Based On Fuzzy Theory

Grid workflow (GWF) has become a basic service grid environment,it can will resources through the workflow (WF) format integration, so asto make the grid users need not concern low-level details. GWF will be alot of users use, that is a key performance. Among them, GWF taskscheduling and load balance is one of the important factors affect systemperformance. This topic research aim is to: solve GWF system2scheduling problem, namely Web Service (WS) mirror choose schedulingproblem.In order to adapt to the loose coupling distributed environment andheterogeneous (each node hardware configuration software, such as theuse of differences between the different performance CPU or operatingsystem, and so on) characteristics, this paper use JAVA language andrealize the WSDM, distributed scheduling algorithm in middlewarestructure will be realized. This can be deployed in separate middlewareservice nodes, and middleware distributed job scheduling, WS node loadconditions the function such as collecting wrapped in WS, more choice tosolve the WS mirror scheduling problem, this paper proposes a kind ofbased on the theory of fuzzy dynamic scheduling algorithm, thetraditional control theory of fuzzy logic control technology is applied toGWF load balance, the algorithm is also looking to the task priorityproblem.We also for GWF system, design WS scheduling middleware （WSDispatch middleware, WSDM） to realize the algorithm. And givesseveral other algorithms are compared with experimental results. Theresults show that, based on the theory of fuzzy dynamic schedulingalgorithm is effective to reduce the GWF system call WS the averagetime, make the WS mirror group of computing resources get morereasonable use.