The following is an excerpt taken from my honours project research proposal.
Grid computing has opened up possibilities for e-Scientists to conduct and collaborate on computer intensive experiments which would have once been infeasible[5]. The next generation of large scale experiments for E-Science requires access to large scale computing resources and data storage, and grid computing has made this possible. High-performance computing experiments can now be run without requiring direct access to a single super-computer.
These experiments are often represented by scientific workflows[19], which involve the break down of an experiment into logical and ordered components that may be responsible for collecting and processing data. These components may be dependent or a dependency for other components within the workflow. These scientific workflows have simplified the process of designing and executing scientific experiments.
Due to the nature of the computing landscape, grids commonly consist of heterogeneous resources; every resource on a grid can potentially have different physical characteristics and a different configuration. For an e-Scientist to successfully use the full potential of a grid they must tailor their experiment to run on all or a subset of these resources. In most cases an e-Scientist may have some experience in software development. However, their main concern is in their field of research. For e-scientists, the process of developing and deploying software across a range of platforms, configurations and organisational boundaries is challenging[6, 7].
One approach to reducing the effort required for developing grid applications is using virtualisation to abstract resource characteristics and allow e-Scientists to define their own run-time environment for an experiment application[3]. Using this approach removes potential application development issues such as portability from the e-Scientist’s responsibility. This can be achieved by using platform virtual machines, which emulate a complete machine including its hardware, operating system, and software. However this method still poses some problems for e-Scientists as the configuration of such environments can be time consuming and requires knowledge of operating systems concepts and system administration.
The complete research proposal can be found here.