Task management driven simulation of grid applications using SimGrid
Narayanappa, Uday Kumar B.
Task graph models are normally used to represent the organization of grid applications in terms of the interaction between tasks. However, in a grid environment there is neither an efficient task management scheme nor any efficient structures to support the execution of large complex applications. This thesis proposes a simulation model that integrates a task management scheme and the libraries provided by SimGrid simulation tool to evaluate the performance of a complex application submitted for execution to a grid-computing environment under a given scheduling heuristic scheme. The efficient structures are used by the task management scheme to monitor the state of execution of completed tasks. The task management scheme uses an alternative representation to the task graph model referred to as the functional description of the application. The proposed simulation model takes the functional description of a complex application as input and outputs the total execution time of the complex application under a given scheduling heuristic scheme. Different task scheduling heuristic schemes are analyzed in this thesis, using SimGrid as the simulation engine. The results reported compare the performance of a complex application under different scheduling heuristic schemes.