The Study And Implementation Of Resources Scheduling Of Collaborative Task In Grid

As a key technology for the next generation Internet, the main target of the grid which linkes geographical computing resoures, storage resources, network resources, software resources, information resources is providing computing and information services for users as a supercomputer. In order to improve the performance of the grid, reduce the execution time and the consumption of the grid computation, some resources scheduling algorithms such as Min-Min, General Algorithm, Artificial Intelligence etc have been presented. At present, the research of algorithm mainly focuses on the resources scheduling of meta-tasks.However, the task is various in grid environment, the task can be divided into sequential task, parallel task, collaborative task, according to the relations between sub-tasks. In grid, the most popular duty is the collaborative task whose sub-tasks have complex restriction relations. The existed scheduling algorithms which are established in the certain supposition have some limitations and the characteristics of grid increases the difficulty of resources scheduling.In view of this kind of situation, a new resource scheduling algorithm-Dynamic Multi Value Scheduling Algorithm of collaborative task is proposed in this article. In DMVSA, the critical path node, the earliest start time and the least start time of every node, the distance between node and critical path are dynamically computed according to the DAG of the collaborative task firstly, the priority of every ready node are calculated based on the above parameters secondly, finally the alogorithm selects resources according to the priority of the node, and the above parameters are calculated according to the scheduling result each time. The paper also introduces the mathematic referent model and the detailed steps and of DMVSA and the computing method of each parameter.In order to confirm the performance of the DMVSA, the simulation is carried on through simgrid simulator. The constitution and the modules of the simulated model of DMVSA, as well as the structure, the characteristic and fuction of each module are discussed detailly in this paper.Finally this paper analysises the time complex degree of DMVSA, DAG-Min and DCPN algorithms which is o(mn~2), and the result of the simulation which is carried on in various computation communication ratio and different coupling rate in view of every kind of collaborative task shows that the DMVSA have the better performance and the shorter finished time Compared with DAG-Min and DCPN algorithms, can't be affected by the computation communication ratio and different coupling rate and have the better compatibility.