Pattern/Operator Based Problem Solving Environments

Abstract

Problem Solving Environments (PSEs) provide a collection of tools for composition of scientific applications. Such environments are often based on graphical interfaces that enable components to be combined, and in some cases, subsequently scheduled on computational resources. A novel approach for extending such environments with Design Patterns and Operators is described -as a way to better manipulate the available components -and subsequently manage their execution. Users make use of these additional abstractions by first deploying 'Structural Patterns' and by refining these through 'Structural Operators'. 'Behavioural Patterns' may then be used to define the control and data flows between components -subsequent use of 'Behavioural Operators' manage the final configuration for execution control and dynamic reconfiguration purposes. We demonstrate the implementation of these Patterns and Operators using Triana [14] and the Distributed Resource Management Application (DRMAA) API [10].