Fault Tolerance Framework for Composite Web Services

Abstract

A composite Web service combines multiple, logically interrelated services for creating more common services meeting complex requirements from users. The services participating in a composition coordinate the actions of distributed activity using Web services protocols to reach consistent agreement on the outcome of joint operation. However, as services run over unreliable protocols, there is a great chance that services fail due to the failure of protocols. However, current protocol standards provide fault-tolerance but are limited to backward recovery using expensive compensation and roll-back strategies. This paper gives an extension of the existing Web services business activity (WS-BA) protocol to deal with failures using forward recovery approach. A set of common failure types affecting the execution of component services is identified, and recovery solutions for each identified failure are also presented. The fault-handling extension of the WS-BA protocol implements recovery solutions for each of the identified failures to handle failures at runtime. Another important aspect about which the WS-BA protocol specification is unclear is reaching and notifying consistent outcome on the completion of joint work. This study extends the WS-BA protocol to notify consistent outcome reached by all participating services. The implementation and testing of the framework are performed using the model-checking and verification tool UPPAAL. A well-known application example supports the study. The key properties of the framework, like the execution of corresponding recovery actions in cases of failures and reaching a consistent agreement on the outcome of joint operation, are verified.

Keywords

• Web services
• fault handling
• forward recovery
• model-checking
• transaction protocols