The Research Of Grid Resource Scheduling Algorithms With Uncertainty QoS Based On Vague Sets

With the development of science and technology and information explosion, People usually need to carry on some large-scale complex computations in computer science, engineering and business fields, more and more high demand of computing performance been proposed. Super computer system as an effective solution of high performance computing, Because of the expensive cost, many small and medium-sized enterprises and scientific research institutes which have real demand step back. In 1998 the Ian Foster Professor, a senior scientist at Argonne National Laboratory in USA, who proposed a new parallel computing architecture for distributed system, it’s the grid. Its goal is to achieve the network interconnection and resource comprehensive sharing, Through the network and protocol, the computing resources, storage resources and information resources that distributed in the global will be connect in together, to form a virtual super computer.It provide low cost and high performance computing services to users through a unified access interface.The grid will become an important development direction of high performance computing.Compared to the traditional distributed computing system, the grid scale is larger, better compatibility of resources.The most important is the capacity of existing resources integration and utilization,It’s able to provide convenient, reliable and low cost resource services to users in the premise of increasing new computing devices, And grid system has the advantages of open, cross platform and easy to expand etc.,The Grid computer technology has garnered wide attention in both academia and industry.Grid resource management and tasks scheduling is the core component of the grid syetem, It’s also an important prerequisite to provide grid resource services and complete user tasks. In the existing grid resource scheduling algorithm that make the quality of service as the goal, most of them only selected an dimension QoS parameters, for less attention on multidimensional QoS, and the scheduling algorithms are based on the accurate matching, there is lack consideration of providing the precise QoS demend difficulty when users access the grid services. In addition, the scheduling algorithms based on accurate matching will cause some problem such as the a smaller resources selection range and the low matching degree. Considering the shortcomings in the existing research, the following works have been done in this paper:The first, the Vague set is used to describe the uncertainty task QoS requirements in this paper, and give the QoS measurement model based on Vague set of accurate values, interval values and qualitative values. We get the Vague value distribution of the upper generalized attribute on each evaluation grade by using the evidential reasoning (ER) algorithm to conduct the information fusion for hierarchical QoS parameters.To calculate the comprehensive utility value of each user tasks, and considering it as a priority of task scheduling.The paper apply Vague set to express the users’real QoS requirements more accurately, and lay the foundation for grid task scheduling.Second,The hierarchical grid resource discovery model based on the multiattributes weighted clustering algorithm is proposed, Through the clustering analysis, the grid resources which have similar performance gathered together to form the resource domain. It provide the integral services to improve the "visibility" of the low connection degree resource nodes in grid, at the same time, It is the parallel way that the grid resource discovery is limited in the local space, and improve the efficiency of discovery and resource utilization.Last, There is a low matching degree problem about the existing grid resource scheduling algorithm based on the grid quality of service, This paper proposed multidimensional QoS grid resource scheduling algorithm based on the similarity of Vague sets, It is moderate mathing and the high QoS demand task priority scheduling, and it provide the resource negotiation mechanism in the acceptable threshold range of user.Through the simulation experiment, The algorithm has better performance in the task average completion time, resource load balancing and the success rate of resources request than min-min algorithm. The algorithm has the better adaptability in "poor" grid environment. It can achieve a better mapping and matching between tasks and resources in the promise of ensuring the user QoS requirments.