The Research On Grid Task Scheduling Models And Algorithms Based On QoS

The grid research, in the current international, is a hot-spot and frontal field for its huge implementation. According to the implement technology of grid system and the definition of the grid, we can conclude that high powered grid task scheduling algorithm is a key technology for the grid, and grid Quality of Service is an important factor which must be considered in the process of task scheduling. Therefore the researching of grid scheduling model based on quality of service and task scheduling algorithm is not only a main contents but also a key contents for the grid .Three grid task scheduling models proposed and several task scheduling algorithms improved based on this proposed models.①Trust-Driven Grid Task Scheduling Model. Trust is also a QoS requirement. Adopting the method of linear compensation calculates the grid trust value. Several kinds of task scheduling algorithms have improved into taking trust as the only scheduling goal. Through the emulation result we concluded that the improved Max-min algorithm(TD-Max-Min) is better then the improved Min-min algorithm(TD-Min-min) as the increase of the strong related task's number. Simultaneity, proposed a double match scheduling algorithm which changed the single goal into a double goal based on the former model. The purpose of this change is to optimiz the trust goal and Makespan at the same time. The emulation result shows that TD-BM is better than TD-Min-mm.②Grid Task Scheduling Model Based on ER Theory. The QoS' value is calculated by ER theory which, compared with the linear compensating method, can deal with the uncertainty information. Proposed a priority and multiple QoS restrained task scheduling algorithm on the basis of ER theory. Compared with the traditional scheduling algorithms, it can shorten the Makespan.③Grid Task Scheduling Model Based on Connective-number. Using connective number of Set Pair Analysis(SPA)to express and process the synthetic uncertainty in task scheduling of the grid. After introducing SPA and its application briefly, present the definition of connective number, operational rules and total order relation suitable for computing grid scheduling. Several algorithms are presented for the uncertain grid. Theory analysis and numerical experiment illustrate that these algorithms can express the dynamics and uncertainty of expected time to compute of tasks in the computing grid environment. Simultaneity, connective number Used to describe the QoS parameters. Task scheduling function was Proposed. The calculation of the QoS' value also adopted the method of linear compensation. However, it is need a further research in comparing of each algorithms performance.