Abstract

This paper examines exception handling in workflow systems. We propose a layered and clustered architecture in exception handling. To detect exceptional situations, we use an approach based on software supervision to compare the observed system behavior with those derivable behaviors in the belief set constructed through default evaluation. We adopt an exception handling mechanism that involves exception masking and propagation. We propose a case-based management system to learn from experience that provides a system assisted decision-making method to derive acceptable exception handlers. Throughout the paper, we use the METEOR workflow system, specifically its workflow model and system architecture, as a concrete example system to explain our approach.