Resource Management And Task Scheduling In Grid Computing Pool

With the fast development of computing technology and network technology, most large-scale science computation applications cannot be accomplished on one single high performance computer. It has to be performed on a super virtual networked computing environment that integrates various kinds of computing systems. Grid computing technology can realize resource sharing easily. In a computing grid, the shared resources are computing resources. A computing grid can integrate all shareable computing resource in Internet into one virtual supercomputer, and provides uniform access interface for users.The research center of power system automation at UESTC is conducting the research on applying grid computing technology for electric power system applications. In this project, an efficient distributed resource management model — computing pool is applied. Computing pool aims at providing unified and transparent service to users by organizing all the computing resource on the Internet in an efficient manner and hiding the internal heterogeneity and dynamics of the computing nodes. It can provide coordinated computing for large-scale scientific computing tasks.This thesis mainly describes the design and implementation of two functional modules: resource management module and task scheduling module.In the design of resource management system, varieties of resources are supervised through LDAP. LDAP is a kind of light directory access protocol, which optimizes the data-read operation. It is suitable for the resource management in grid service that requires frequent reading operations.When designing the task-scheduling algorithm, the Adaptive Genetic Algorithm was selected. AGA is a kind of bionic optimization algorithm based on Genetic Algorithm. The advantage of AGA is that the algorithm is independent of the particular problem or the object to be analyzed. All it needs is the performance's measure of the system. Good performance is achieved by using AGA in the scheduling of tasks in the grid computing pool.