Self-similarity and multidimensionality: Tools for performance modelling of distributed infrastructure
Chapter in Scopus
- Additional Document Info
- View All
In this article, we present a model that uses large-deviations and the Pareto probability distribution to model self-similar data in high-performance infrastructure, such as the one found on computational and data grids, transactional and computational clusters, and multimedia streaming. We also show how Principal Component Analysis can reduce dimensionality of data, such as the one produced by different types of problems, user preferences and behaviour, and resource popularity. © 2008 Springer Berlin Heidelberg.