Planning for stream processing systems

Abstract

With the advent of compositional programming models in computer science, applying planning technologies to automatically build workflows for solving large and complex problems in such a paradigm becomes not only technically appealing but also feasible approach. The application areas that will benefit from automatic composition include, among others, Web services, Grid computing and stream processing systems. Although the classical planning formalism is expressive enough to describe planning problems that arise in a large variety of different applications, it can pose significant limitations on planner performance in compositional applications, in particular, in stream processing systems. In this paper we extend the classical planning formalism by introducing new language constructs that support the structure of stream processing domains. Exposing this structure to the planner can result in dramatic performance improvements: our experiments show exponential planning time reduction in comparison to most recent metric planners. Copyright © 2005, American Association for Artificial Intelligence (www.aaai.org). All rights reserved.