| Grid computing is a new distributed computing technology which builds on Internet. It can provide the members of dynamic changing organizations with wider sharing resources. Resource allocation strategy and task scheduling algorithm are both important components of grid computing. But either resource allocation or task scheduling is a NP-complete problem, so they become a research focus in grid computing area.In this paper the architecture of SOA-based enterprise grid computing system is detailedly researched, the resource allocation strategies of resource management as well as the scheduling strategies of load management in current enterprise grid are analzed and summarized. For the structure of resource management, a resource optimization strategy which based on reservation and preference is proposed, namely, selecting resource that meets QoS of tasks by reservation principle, and selecting the resources that have successful scheduling experience by preference principle. For the structure of load management, build a tabu list by obtaining the optimization result of resource management; the algorithm of this paper improve the generation of initial population and the mutation, enhance the performance of scheduling by building the proper fitness function and controlling other parameters.The theoretical analysis and the experiments show that the allocation strategies and the scheduling algorithm proposed in this paper not only meet the tasks QoS but also overcome the shortcomings of traditional genetic algorithm, meanwhile shorten the total execution time of enterprise grid.It's just a preliminary research for allocation strategy and scheduling algorithms of enterprise grid in this paper. The tabu list is too great will become the bottleneck of system performance, which still needs improvement. Moreover just the genetic-tabu algorithm applies to the scheduling in this paper, for the application of other algorithms still needs further research. |
PDF Full Text Request