Composite web services can be orchestrated in a decentralized manner by breaking down the original service specification into a set of partitions and executing them on a distributed infrastructure. The infrastructure consists of multiple service engines communicating with each other over asynchronous messaging. Decentralized orchestration yields performance benefits by exploiting concurrency and reducing the data on the network. Further, decentralized orchestration may be necessary to orchestrate certain composite web services due to privacy and data flow constraints. However, decentralized orchestration also results in additional complexity due to absence of a centralized global state, and overlapping or different life cycles of the various partitions. This makes handling of faults arising from composite service partitions or from the failure of component web services, a challenging task. In this paper we propose a mechanism for handling faults in decentralized orchestration of composite web services. The mechanism includes a strategy for placement of fault handlers and compensation handlers, and schemes for fault propagation and fault recovery. The mechanism is designed to maintain the semantics of the original specification while ensuring minimal overheads.
展开▼
