Design And Implementation Of A Grid Simulation Framework Using GridSim

Grid Computing is now a mainstream technology of parallel and distributed computing. It is an emerging technology that focuses on uniformly aggregating and sharing distributed heterogeneous resources for solving large-scale applications in science, engineering and commerce. Because of the special features of the dynamic heterogeneous grid environment, the researchers hardly evaluate performance of lots of grid systems. The access grid nodes are not widespread. Therefore, more and more researchers use grid simulators for implementing, testing, and evaluating their ideas in resource management and scheduling algorithms.Motivation and goal of this thesis is providing better simulation environments that allow researchers to focus on their theoretical research. After analyzing grid simulation tools in depth that currently exist, an idear that use XML to describe grid resources, tasks and user's QoS is proposed. Then, the GridSim Simulation Framework (GSSF) is presented. It provides a XML parser module to map XML description to GridSim entity. GSSF also provides a grid scheduler module to enables easy-to-use experimental studies of various grid and local level scheduling algorithms. In order to allow users to describe tasks and their dependencies, a simple and flexible XML-based GridSim Workflow Description Language (GSWDL) is defined. Then, a distributed scheduling workflow simulation engine (WFSE) together with GSWDL is designed. A message-driven scheduling architecture for managing dependence among tasks and communicating between tasks is shown in WFSE. In order to use GSSF easily, a grid generator called GridSimG that built on GridG and GWA is proposed. It is written in Python 3.0. GridSimG is used to generate XML-based resources and tasks.Applications show that GSSF provides an easy-to-use and extensible grid simulation environment. It makes a good foundation for the future research in grid simulation.