DEE: A Distributed Fault Tolerant Workflow Enactment Engine for Grid Computing

It is a large and complex task to design and implement a workflow management system that supports scalable executions of largescale scientific workflows in distributed and unstable Grid environments. In this paper we describe the Distributed workflow Enactment Engine (DEE) of the ASKALON application development environment for Grid computing. DEE proposes a de-centralized architecture that simplifies and reduces the overhead for managing large workflows through partitioning, improved data locality, and reduced workflow-level checkpointing overhead. We propose a systematic approach to understanding the nature of the fault tolerance and enactment engine overheads that impact the distributed Grid execution of workflow applications. We report experimental results for a real-world material science workflow application.