Motivations and Goal
The advent of parallel scientific computing on multi-cluster distributed platforms that span multiple organizations, often termed “grid computing”, over the last decade has generated tremendous research activity as several applications can take advantage of such platforms. In most of these researches, having a synthetic but realistic representation of the underlying infrastructure is often useful.
The goal of the Platform Description Archive is to provide a place where researchers can exchange descriptions of various large scale platform configurations. To ease such exchanges we aim at providing a description formalism, generation and extraction tools and a database of currently deployed grid computing platforms.
Approach
Our approach to build the Platform Description Archive is:
- Define a platform description XML formalism that is:
- Readable as the platform description files should be easy to parse by computers but also easy to read or modify by human beings. This should ease the exchange of files between researchers.
- Reusable as it should be possible to use the same platform description file to run several experiments.
- Expressive as the formalism should allow to declare additional arbitrary data that can be useful to end users.
- Compact as platform description files should be small enough
(up to hundreds of KB) to allow the generation of multiple
files representing a large range of scenarios in a reasonable
space.
Note a full description of the proposed formalism is available here.
- Give access to the description of several currently deployed grid computing platforms.
- Provide tools to randomly generate platform configurations and to extract subsets with given characteristics from a larger platform.
- Provide conversion scripts to and from other description formalisms.