Research Of Constrained Multi-objective Grid Task Scheduling Model

Grid is a new way of computing. Since its ability to share, choose and cooperate withheterogeneous resources located all around the world which can be used to solve the largescale computing problem on science, engineering and business, it is researched by more andmore people. To make grid become the mainstream computing mode, there must be aneffective way to solve the safety problem brought by the dynamisms and uncertainty of thegrid environment and meanwhile it also can make the users who have free resources distributetheir resources. One of the better ways is to introduce the reliability and economic factor togrid environment and then build up and manage the grid resources.By dispatching the user tasks through researching the grid system and submitting toprocessing carried out by different resources, this paper introduces the Simulated SchedulingAlgorithm which is both reliable and economic. This method assesses the reliability andcomputing costs of entity behavior which can reflects the reliability and cost performance ofthe entity grid behavior and achieves the overall goals of maximal economic benefit. Thispaper increases the effectiveness of the data needed in the task scheduling in grid and theauthenticity of the scheduling results by using the more common simulation tool GridSim tomake the initial configuration on the environment. This paper has the following main jobs:（1）It will dig into the basic principles of Grid Task Scheduling strategy and comparesome different task Scheduling strategies; analyzing its advantages and disadvantages andpropose an improvement solution; maximum the total utility value got by the grid systemand the economic benefits got by the users meanwhile its reliability is guaranteed.（2）By digging into the basic principles of Ant Colony Algorithm, it will make full use ofthe positive feedback system, combine the Scheduling Model with Ant Colony Algorithm,and introduce the Local pheromone update strategy, which quickens the convergence rate ofalgorithm.（3）Combining the simulation tool GridSim, it will realize the Grid simulation platformon which the researchers can configure the grid, simulate the environment, call the task scheduling algorithm, and verify its feasibility and performance. At the same time, thisplatform offers an interface where the researchers can expand their own task schedulingalgorithm. It brings convenience to them and meanwhile reduces their cost on learning grid.