Grid Resource Assigning And Task Scheduling Algorithm Based On DBC

Many problems must be solved in order to implement efficiently Grid computing, and how to efficiently manage resource and schedule computation in grid environment is one of the most elements which may affect the success of grid computing. With the study by some scholars, they found introduce market economy concept to the grid management and design is a good choice.Time and cost of the task scheduling are most important QoS factors in the market economy. Based on these two factors, DBC scheduling strategy was proposed by Dr. R.B. There are three heuristic algorithms in the policy; each has its target to optimize including time and cost. However, these algorithms have some weakness, such as all tasks may be intently assigned to only a resource, then caused the resources load not to be balanced. In the actual application, the user hope the task will be completed most maximally under the time and expenditure condition, but these three algorithms have not taken optimized by the tasks completion factor. In the process of scheduling and carrying out all the tasks of the users, users’ QoS factors are static, these three algorithms had not considered the situation of time and budget are dynamic change. However, in the real market economy, the price budget and the time deadline often are the dynamic change.On one hand, in view of the above shortcomings, this paper present a multiuser task scheduling model based on the situation that consider time and budget are dynamic change from the very beginning(Along with task execution getting smaller, just started to change is quite quick, afterward changed is quite slow. and the task execution priority has been considered),and a improving grid task scheduling algorithm based on Budget and Deadline benefit function is proposed, and running the algorithm onto the simulator Gridsim. This algorithm tallies with the actual situation more. The comparison results show that we can gain more completion tasks and better performance of load balancing by utilizing the algorithm presented in this paper.On the other hand, particle swarm optimization algorithm is a biological modelling evolution computational method which presented by Eberhart and Kennedy in1995.It is a value powerful tool for searching the optimum value of a numerical function in the continual territory. At present, the researches on particle swarm optimization algorithm which used in solving discrete questions also has reached good achievement, and grid resource assigning and task scheduling is a very typical discrete space optimized question. Therefore, the particle’s position and velocity was redefined,and the switch rules of particle’s position and velocity was redesigned. At last, this paper presents two improved discrete particle swarm optimization algorithm by unifying discrete particle swarm optimization algorithm with the simulation annealing algorithm, the Stretch function, and the genetic algorithm. Then, using the improved algorithm to search the optimum plan about minimizing the actual executing time and cost under the budget and time limit, and get the better systems synthesis target. In this paper, we run the algorithms onto the simulator Gridsim. The simulated experiments indicate that the methods presented in this paper can produce good results.