Research On Grid Resource Scheduling Algorithm Based On Econometrics Model

Grid computing is the next-generation computing platform to address the issues raised by large-scale resource-intensive. It is a development direction about the parallel and distributed processing technology. And the computational grid resource management is one of the key technologies. However, because of the distribution, heterogeneous and dynamic nature of grid resource management, it is more complex than in distributed computing environment and cluster computing environment. It is not only to support task scheduling, real-time monitoring of resources and job execution status in inter-organizational or management domain, but also to protect the local site autonomy, and provide the appropriate QoS support. In grid environment, how can scheduling the resources effectively is a very important issue.In the study of applying market mechanism to grid resource allocation, an influential prototype system is Nimrod/G which is developed by Rajkumar Buyya in Australia's Monash University. Rajkumar Buyya and others discussed some grid scheduling algorithms developed by them constrained by the task deadline and the budget of cost (Deadline and Budget Constrained, DBC), specifically including the cost optimal scheduling, time optimal scheduling, and the cost - time to balance scheduling.In this paper, the resource scheduling algorithm which based on the econometrics model was analyzed extensively. By this model, the resource owner can offer and upgrade sharing resources more actively. In addition, this model is more convenient to use resources. At the same time, based on the characters of grid resources, this paper made some improvement on the primary algorithms, which are the resources scheduling algorithms based on Deadline Budget Constrain. And the improved algorithm can satisfy the users'need on completing more jobs. The main content of the article includes 4 aspects as below:(1) Summarize the research state and features of the grid resource scheduling systems.(2) Analysis some important models and algorithms in grid resources scheduling, and focus on the algorithms based on Deadline and Budget Constrained in the econometrics model.(3) Improve the primary algorithms and propose a new algorithm called budget/deadline algorithm to meet the users'requirements of task output.(4) Carry out a simulation experiment on improved algorithm by grid simulator. And the experimental results are given to show that the proposed algorithm increases the task output greatly and give better cost-effective performance.