High-performance computing for eScience with iPSE approach: principles, implementation and applications

Abstract
The iPSE (intelligent PSE — Problem Solving Environment) approach allows constructing composite applications for computer modeling using existing software as remote computation services for solving particular subtasks. To provide users with advanced software tools (with such quality requirements as computing performance, precision of the result and usability) it involves composition of expert knowledge bases, artificial intelligent technologies, rich user interface etc. One of the most significant features of the approach is tight user interaction within virtual professional society for knowledge exchange and collaboration. Based on service-oriented architecture, this approach proposes methods for service integration irrespective of software and hardware platform, access protocols or algorithm traits. Thus the iPSE approach allows utilization of different infrastructure architectures including set of computational clusters, Grid etc. The talk will be focused on the basic principles of the approach. Some of them are already implemented in practice; some are concerned to be implemented within ongoing projects. Selected experience of eScience Research Institute, acquired during number of projects in various areas of knowledge (e.g. quantum chemistry, hydrometeorology, complex system simulation etc.), will be shown.