The continuous evolution of computing and networking technologies makes available to an increasing community parallel and distributed systems that can deliver impressive performance, along with a large amount of software, data and other resources. However, using such resources in an efficient way is generally assessed to be a very hard task. The development of efficient software requires a deep knowledge of the architectures at hand and a consistent programming effort to deal with architecture-related details, such as memory hierarchy access, process decomposition, scheduling and mapping, inter-processor and inter-system communication implementation, and load balancing. Furthermore, the different features of highperformance hardware/software platforms prevent from an easy, efficient and reliable reuse of existing high-quality software. Many other difficulties arise in the context of Grid computing, to enable flexible, secure and coordinated sharing and coupling of distributed heterogeneous resources. Therefore, the user community strongly recognizes the need for advanced environments, which support the level of abstraction required by an effective development of complex high-performance applications and preserve, in the meanwhile, properties such as reliability and performance. Such environments should also allow users to easy integrate the many software tools required to develop a large and complex application (supporting libraries, highly optimized computational kernels, tools for data visualization and analysis, etc.). Different ''solutions'' have been proposed, in terms of programming paradigms, computational methodologies, software architectures and corresponding implementations, to provide users with suitable, high-performance, advanced programming environments. As an example, algorithmic skeletons, object-oriented frameworks, pattern languages and component-based systems have all been recently considered in this perspective. Nevertheless, despite active research is currently being performed in the field, many problems still need a solution.
This special issue contains most of the papers presented at a minisymposium that took place within the Parallel Computing 2001--ParCo 2001 Conference (Naples, September 2001) and other papers from authors actively and successfully working on representative projects related to the topics outlined above. They discuss problems, methodologies, techniques, tools, as well as proposals and implementations www.elsevier.com/locate/parco