Design And Research Of Scheduling Models In Economic Grid Based On Gridsim

Grid, emerging as a new paradigm for next-generation computing, enables the sharing, selection and aggregation of geographically distributed heterogeneous resources for solving large-scale problems in science, engineering and commerce. Availability, usage and cost policies vary depending on the particular user, time, priorities and goals. The management of resources and application scheduling in a large-scale distributed environment is a complex task.This thesis provides a new scheduling algorithm based on granularity and combined tasks by analyze the process of scheduling users jobs and dispatch them to the various resources in Grid. This method combines several small-grained tasks into a group to submit them to the appropriate resource. Finally, a popular simulator in economic Grid called Gridsim is used to simulate, compare and analyze the implementation by different parameters. And the results show the good effect of this algorithm which is to support the future research work. And it analyzes the simple and efficient methods of deciding the winning bid in all the buyers and sellers of the various auction models in economic Grid, and provides a new improved genetic algorithm by adding deteriorative solutions with a certain probability. This algorithm rectifies the defect of early convergence in traditional genetic algorithms and jumps out of the local optimization to find out the overall optimization to the full. At the same time, convergence of the algorithm is ensured. Finally, the algorithm is simulated and implemented by coding. And the result shows the simplicity and efficiency of the process to get or approach the optimization, and it lays the foundation for the next step of study work of auction simulation in economic Grid.Finally, it is researching the implementation schema of combinatorial double auction in economic Grid by providing two different kinds of combinatorial double auction models which are based on central-proxy mode and distribution mode separately. In central-proxy mode, bid information is not known for the users in Grid. A central-proxy is used for the process of pricing and task distribution. And in distribution mode, Contrarily, bid information is all known for the users in Grid. So a central-proxy is not needed, and instead user nodes implement these functions. Besides, an improved genetic algorithm by adding deteriorative solutions with a certain probability is used to decide the winning bid for the auction. Finally, experiment of the combinatorial double auction of two different modes is implemented by the simulation tool Gridsim. And the data results validate the effectiveness of the combinatorial double auction based on these two modes, also it can be extended well to the more fields.