Decentralized orchestration of composite web services
Proceedings of the 13th international World Wide Web conference on Alternate track papers & posters
Handling faults in decentralized orchestration of composite web services
ICSOC'05 Proceedings of the Third international conference on Service-Oriented Computing
| Hi-index | 0.00 |
Recently there has been an increasing interest in workflow correctness and failure handling requirements, mostly in the context of centralized workflow control architectures. Centralized workflow control however suffers from reliability and scalability problems due to the use of a central engine to manage state information of all workflows in progress. A distributed workflow control architecture on the other hand offers much better reliability and scalability. However, developing mechanisms for handling correctness and failure requirements for distributed workflow control requires careful consideration of a number of pragmatic issues. In this paper we discuss some of these issues based on the experience gained while designing a rule-based prototype distributed workflow management system. Agents in distributed workflow control are responsible for both step execution and managing workflow state information. We first classify agents into different types and discuss the functionality to be provided by each type of agent. We then discuss how events are to be propagated between the agents as efficiently as possible while still satisfying the failure handling and coordinated execution requirements. Finally, we present a detailed discussion on performance issues. We analyze the performance of a distributed workflow control system during normal execution and when special correctness/failure handling requirements are to be satisfied. We also compare and contrast the performance of a distributed workflow control architecture with respect to central and parallel workflow control architectures.