Exception Handling Mechanism Based On Multi-agent Virtual Organization Workflow Management System

The importance of exception handling is well recognized by system designers and software engineers. Workflows can be characterized as sequences of program invocations and data exchanges between distributed and heterogeneous stand-alone systems. Workflow Management System (WFMS) is tool and platform for designing and executing of Workflow. For special environments and long duration of executing, there are variances of possibilities to fail when Workflow running. A mechanism is necessary to maintain WFMS' robustness while error or failure happens. This paper introduces our Agents-Based Framework for Software Composition and WFMS implemented with it, including framework structure and system execution. Then we thoroughly describe the structural exception handling of WFMS: Possibilities of exception are defined as Concept Model in system shared Ontology; Exception instances detected by Negotiation, Communication and Domain Operation are stored in Run Context of Activity Instance; Exception Handling as entities of Agent Model can be defined, reused and inherited by system and user. According to topo-structure of Activities dispatch, exception instances can be transferred between Agents to search one that can handle such exception. Last, we discuss Transaction Model and Failure Recovery of WFMS, propose General Resource concept and methods of implementing resource-centric transaction model, integrate the plug-in container of Resource in ABFSC architecture, separate Resource scheduling as independent functional module from Skill module, and imply Transaction planning in Agent Collaboration, without losing flexibility and interaction ability of ABFSC architecture. Thus provide the support of concurrency and recovery for shared data Resource in Workflow, which is the execution basis of data consistency control and exception handling.