Research On Resource Management And Scheduling Based On Virtual Organization In Grids

Grid Computing is currently a hot topic in Internet research, but also a development direction of parallel and distributed processing technology. It realized the comprehensive sharing of computing resources, data resources, storage resources, information resources, knowledge resources and expert resources by integrating the entire network into a single giant super-computer and utilizing the idle processing power of the web. In grid computing, resource management, task scheduling and task management are the three basic functions that the grid must have. In this paper, a resource management based on virtual organizations has been studied and task scheduling in conjunction with the market economy model has been explored, the main research work and contribution are shown as follows:The resource sharing and collaborative working mechanisms and theoretical models in the VO-based grid computing environment are discussed in this paper. Through the study of VO, a new link bandwidth collection model of the VO and a simple and effective algorithm for VO aggregation has been proposed. A Hash Table-Binary Sort Tree (HT-BST) resource management model has been proposed for the dynamic change of the resource properties and the resource update operation is much higher than the operation of Resource Register. The model takes into account the strengths and weaknesses of the existing list structure. The registration resources are linked by ID in order through the BST structure to achieve the optimization efficiency of resource discovery and search, and to fully meet the rapidly changing characteristics of the network load and various properties. A simple and effective resource accounting model has been proposed by using the price of leverage in the market economy, and it implemented the load balancing of resource within the VO.This paper described the architecture of Nimrod-G and the entity's work flow within Nimrod-G which was the resource broker based on the Grid Architecture for Computational Economy developed by the Australian Monash University. It also described DBC time optimal and DBC cost optimal scheduling algorithm of Nimrod-G. The LP-DBC optimal time and the LP-DBC optimal cost scheduling algorithm have been proposed on the basis of the existing algorithms. The algorithms is based on a new idea, the task scheduling problem was converted into a linear programming problem, each resource gains the optimal execution time and the optimal number of task allocation and the smallest completion time in the use of the simplex algorithm to obtain the approximate optimal solution.GridSim was chosen as the grid simulation tool, the broker-based grid simulation platform was built, and the graphical user interface client using Java language was implemented in this paper. The simulation results show that the LP-DBC algorithm and the DBC algorithm are basically similar for the task allocation scheme. However, the total task's completion time of LP-DBC algorithm is significantly less than that of DBC algorithm of Nimrod-G. This also shows that LP-DBC algorithm has low time complexity. With the grid task granularity increasing, the simulation results show that LP-DBC algorithm for scheduling tasks in the time overhead getting smaller and smaller, indicating the LP-DBC algorithm has a certain value of using.