Archived: 30 January 2004
University of Adelaide Honours thesis, November 2003.
Supervisors: Paul Coddington and Andrew Wendelborn
The aim of grid computing is to provide an environment which allows easy, ubiquitous access to geographically distributed, heterogeneous computing resources. The effective coordination of such a collection of resources is not a trivial task. The machines comprising the grid and the connections that link them are commonly shared and non-dedicated so their utilization and capacity is constantly fluctuating. Due to the complex concurrent environment presented by a grid the most accurate method of performance evaluation of adaptive grid technologies is observation of actual execution in controlled repeatable experiments. Unfortunately due to the uncontrollable, unpredictable nature of real grid environments, experimentation with grid technologies such as applications and resource schedulers is very difficult. This project implements a powerful grid computing testbed which allows repeatable controllable emulation of arbitrary grid environments in order that grid application and resource scheduler performance may be systematically analysed. This testbed allows unmodified grid applications to run in a highly configurable virtual environment, enabling research into grid application design and grid topology design, both areas which are currently under-researched, as grid computing is a relatively new technology. The testbed will be used to study real grid applications, with a particular focus on application performance in larger metropolitan and wide area networks, as the problems faced by grid application developers are generally compounded in these environments.