Processament d'altes prestacions: arquitectura, entorns de desenvolupament i aplicacions

The objective of this project is to study different aspects of parallel computing, from the processor level to applications. The project is organized in four basic areas: Environment for performance evaluation and applications tuning in parallel and distributed systems. This research covers parallel program development using programming frameworks and templates, including techniques for run-time automatic performance evaluation and improvement of those programs. Strategies and tools for distributed systems management. Four different approaches in parallel/distributed computing are evaluated: Non-dedicated clusters and Grid environments, dedicated cluster based on user-shared machines, geographically distributed dedicated clusters and web-computing. Multiprocessor architecture. The advanced "fetch unit" of high performance processors and the interconnection network behaviour are the selected topics for concentrating our research effort in order to improve the system performance. Applications. Three applications demanding high performance computing offered by the parallel processing are investigated. The first one is related to simulation of complex systems in the field of ecology using the parallel version of the discrete event simulation strategy. The second application is focused on implementation of the video-on-demand (VoD) service, based on distributed systems. The last application is concerns the use of advanced simulation techniques to face the problem of forest fires propagation simulation using parallel/distributed systems.