Research Of Grid Resource Scheduling Based On Multi-objective Decision Makings

Grid computing is one of the most attractive research fields currently, which enables all the geographically distributed resources to be widely shared and collaborate, such as high-speed Internet, mainframes, large-scale databases, sensors, remote devices, etc. Because of its distributiveness, heterogeneity, dynamic, and possessed by many organizations and distinct using, accessing, mode of charge, etc, resource distribution under grid environment is very difficult. Resource distribution has been one of the key research fields of grid systems.Throughout analyzing and researching the resource scheduling and the load balancing of the computational grid system at the present, the defects in the computational grid resource scheduling mechanism thesis are pointed out, such as without thinking about the unique requirement of the users before resource scheduling and the load balance and so on. In order to solve the challenging resource scheduling problem in Computational Grid, this thesis takes the unique requirement of users, resource information and resource load balancing into account, and realizes the resource scheduling for Computational Grid. Research in detail on resource scheduling:1. A XML-based resource representation model and XQuery-based resource querying model are put forward, which have the characteristics of semistructure, briefness and transplant.2. A personal resource requirement description mechanism is put forward, by which computational grid jobs can express their software and hardware related resource requirement on-demand. The mechanism helps accelerate the process of resource seeking and matching, and especially provides powerful support for optimized resource scheduling for different jobs.3. A resource scheduling algorithm is designed and implemented, which takes users requirement and load balance into account. Based on Fuzzy Multi-objective Decision Making, the scheduling algorithm takes users satisfaction as a main target, weighs candidate computational grid nodes, considers the system load and chooses an appropriate node for a job from the candidates finally.The investigations for grid resource scheduling in this thesis consider many kinds of available grid resource conditions, match personal resource requirements, satisfy the demand of users at large and balance system load. This thesis realizes multi-objective grid resource scheduling and improves service quality of grid system.