A generic framework for grid-enabled visualisation and computational steering, and its characterisation

Abstract

This thesis presents a generic framework for grid-enabled visualisation and computational steering that is amenable to characterisation using mathematical models based on benchmarked resources. The models enable designers, providers or users to select the most suited interactive computational steering and rendering resources available to them on the Grid. The framework for visualisation aims at giving a greater performance than a desktop computer can provide, achieved at a cost much cheaper than a CAVE, while allowing simultaneous usage by multiple users. This framework provides multiscale multimodal multiuser grid-enabled visualisation and has the potential to assist the penetration of the Grid in to the domain of advanced interactive 3-d visualisation and geographical rendering from a user’s desktop. While the existing desktop tools and techniques for interactive visualisation are of a general purpose nature and offer limited compute and data intensive graphical visualisation and interactivity, this framework leverages the power of the grid to offer: 1) more advanced visualisation 2) real time interactivity with the rendered content; 3) and integration with key data intensive graphical applications that would benefit from the use of the grid. The framework is conceptual, to allow users to analyse and/or design for their own specific application, but a comprehensive example application has been implemented. To enable the characterisation of a framework for grid-enabled visualisation and computational steering, a mathematical model is constructed to help identify the complicated relationships between the variables of the architectures and systems used. The creation of this model allows for it to be used in a number of ways, from the prediction of application performance on a grid, the comparison of site and grid performance, to suggesting for grid providers. A roadmap is presented for the future evolution of this approach.