The Research On Task-Scheduling Policies Under Grid

The grid technology aims at solving the problem of information islands by connecting all shared network resources and converting them into computing capabilities. Users have to allocate tasks after submitting them to the grid; that is, the allocated tasks will be carried out by the grid resources. Grid scheduling strategy directly determines the grid overall performance scheduling. The most general objective of Grid scheduling is Makespan, which means effective distribution of grid resources in scheduling system. This is commonly suitable for a given task of the potential pool of resources which meet a pre-defined scheduling constraint and from which the appropriate subset of resources are chosen. Finding such a Makespan is a NP-complete problem.In the based on existing grid resource scheduling technologies, the paper used in the grid serverces environment for single processing environment, parallel multi-processing and cacluation of the distribution of resources, Based on the proposed dispatch center from the distribution of tasks model, and in accordance with the grid load,the distribution of tasks for the light load task scheduling and task scheduling heavy load on the different tasks to different task scheduling strategy. To the light load tasks, the users to carry out tasks assigned as a whole, not to break down the tasks, and directly to assign this tasks to an arrangement to meets the mandate boost for implenebtation of computing nodes. To the heavy load tasks, we are classifying the tasks effectively, according to the different computing nodes to sub-tasks for the units to the resources Fenwick scheduling. Through analysis of the effectiveness, of the echeduling strategy is not only optimized globally, but also soved the bottleneck of the dispatch center and local pitimization problem.By analyzing the grid resource scheduling principles, characteristics and architecture, as well as in grid in the resource discovery, registration and inquiry in grid, the paper will apply genetic algorithm in grid scheduing. The encoding parameters, the initial population, fitness function, genetic operator and control parameters to improve the design. Under Gridsim simulation environment, through exeperiments can be genetic algorithm suitable for large-scale grid task scheduling via experiments.