| An organization's systems must respond to changes in the processes they automate. When an organization delivers web services for composition into one or more inter-organizational workflows, their independent evolution can cause problems. Specifically, when previously expected messages between a service and its partners become unexpected due to that service's evolution, the distributed service workflow can fail. This occurs because the respective workflow models of the components are no longer synchronized.; This work presents an intelligent-agent conversation framework with which web services are implemented. The system can adapt web service workflows in response to failures caused by out-of-sync workflow models through the use of a globally shared failure recovery policy. This policy allows agents to resolve various types of model mismatches that cause interaction errors, including changes to required preconditions, partners, and expected message ordering. Case studies are also presented that illustrate the approach of the framework and its assumptions. |
