Optimized P2P Data Management for ReliableWorkflow Execution in Mobile Environments

Abstract

Workflows allow to build complex applications out of existing services. Hence, workflows are inherently distributed. At the same time, workflows are mostly executed by centralized workflow engines that invoke the services of a workflow in a request/reply style. These engines may become a performance bottleneck and limit the scalability of the entire system. As a consequence, approaches to distributed workflow execution have been proposed. While these approaches have better scalability characteristics, they require additional efforts to deal with (partial) failures of the system. To cope with such failures, workflow instance data must be stored redundantly at different sites. This is even more important in mobile environments or mixed mobile/stationary environments where the nodes hosting services for workflows may be mobile and are thus more likely to fail permanently or become temporarily unavailable. This is the case, for instance, for sensor net applications where mobile devices capture data, or when smartphone apps share and jointly process data. In this paper, we present the combination of OSIRIS-SR, an extension to the distributed and decentralized workflow engine OSIRIS that focuses on reliable workflow execution by means of instance data replication, and Compass, an extension to Chord that is particularly tailored for efficient P2P data management on mobile devices.