The Research On Evaluation Model Of Grid Fuzzy QoS And Application

Quality of service is one of the three basic principles of grid, on which the research has not been stopped. Grid QoS management and negotiation system and layered QoS structure were presented, but there has not been a mature method for the computing and evaluation of grid QoS, especially for QoS with potential uncertainty. In addition, grid task scheduling is the key work of grid, and the research on it is stepping into maturity. Not only were heuristic Min-Min algorithm and Opportunistic Load Balancing algorithm (OLB) based grid task static scheduling proposed, but also the dynamic scheduling algorithm presented, including online mode and batch mode. Also the Genetic Algorithms(GA) and Simulated Annealing(SA) algorithm were utilized to process grid task scheduling. But how to better guarantee quality of service during task scheduling and how to efficiently guide task scheduling with grid QoS are still open to be researched.Three evaluation models of grid QoS are proposed and the traditional computation and evaluation of grid QoS is improved.1, Vague set based grid QoS evaluating model. Vague set is used to describe the uncertain express of QoS, such as'around a value'and'security is high', and D-S theory based ER algorithm is used to compute and combine multiple QoS parameters. The result of ER computing is added into task scheduling algorithm and the algorithm is improved. The experimental result has proved that the new task scheduling algorithm guided by Vague set describing QoS reduced the makespan of scheduling and made the grid resources utilized more efficiently.2, Connective number based grid QoS computing model. Connective number of Set Pair Analysis(SPA) is used to cope with the uncertainty existing in grid QoS. The parameter'b'in connective number'a+bi'is to describe the fluctuating range of QoS attribute, and the connective number operational rules and total order relation are studied to compute and evaluate grid QoS, which is combined into task scheduling. Numerical experiment illustrated that being compared with the scheduling algorithm restrained by real number QoS, this scheduling algorithm not only maintained makespan within reasonable area, but also increased the utility of task-resource matching, and enhanced the rate of resource utility.3, Grid QoS evaluating model of hybrid multiple QoS. A new structure of grid QoS with description by real number, interval number and Vague uncertainty is proposed, which embraces all kinds of QoS attributes including logic parameters, network throughput and latency, security, trust, availability, and even the executing time and price. Then the hybrid multiple QoS parameters are calculated and evaluated with ER algorithm, and the application of this composite evaluation in grid task scheduling is introduced.