An adaptable framework for distributed and parallel applications

• Authors: Fernando Rocha;Sérgio Soares;André Soares;Ricardo Lima
• Venue: Proceedings of the 8th workshop on Parallel/High-Performance Object-Oriented Scientific Computing
• Year: 2009

Complex applications need powerful computers to achieve results in a valid time. However, such machines are very expensive and not always available. An alternative is the adoption of several simple, less powerful, and less expensive computers, which triggers the need for an adaptable framework that would support parallel and distributed execution in several environments and with different communication protocols. By executing in a heterogeneous environment, the user benefits from a higher number of machines. The framework we propose allows switching from a heterogeneous to a more specialized executing environment, if or when it is available. It supports both newcomers who can use existing default implementations to more quickly create applications of their own, and experienced programmers who, besides using existing framework abstractions and implementations, can also implement some parts to increase performance. In fact, the main objective is supporting researches, being acceptable to loose some performance when compared to a tailored, and harder to implement solution, still having a better performance when compared to a nonparallel approach.