A knowledge‐based approach for task allocation in heterogeneous distributed computer systems

Abstract

Abstract: This paper presents a knowledge‐based system for task allocation in a network of processors. Since the problem is NP‐complete, optimisation techniques are computationally prohibitive and impractical. Our approach reduces computational burden by using (1) a two‐level hierarchical scheduling model to alleviate the problem complexity, and (2) a search strategy that uses knowledge from the knowledge base to limit the search space to a manageable size. The major goal of our approach is to generate good but not necessarily optimal solutions quickly. The system is implemented on a Sun SPARCstation 1 + using Common Lisp. Our computational experience in using this system for task allocation indicates that the two‐level hierarchical approach generates better schedules in shorter computational time than a non‐hierarchical approach.