Run-time Adaptable Business Process Decentralization

BPEL specified business processes in the Service Oriented Architecture (SOA) are executed by non-scalable centralized orchestration engines. In order to resolve scalability issues, the centralized engines are clustered, which is not a final solution either. Alternatively, several decentralized orchestration engines are being emerged with the purpose of decentralizing a BPEL process into fragments, statically. Fully decentralization of a process into its building activities is an example of static fragmentation methods. The fragments are then encapsulated into run-time components such as agents. There are a number of attitudes towards workflow decentralization; however, only a few of them consider the adaptability of produced fragments with a run-time environment. The run-time adaptability can be studied from different aspects such as the proportionality of workflow fragments with number of machines dedicated to a workflow engine or runtime circumstances such as available bandwidth. In our opinion, the SOA suffers from the lack of decentralization adaptability with run-time environments in the orchestration layer. It demands the mapping of run-time circumstances to a suitable fragmentation model. In this paper, a mapping algorithm is presented, which is based on the number of machines and available bandwidth. Evaluation of the presented algorithm for adaptable decentralization demonstrates an improvement of the bandwidth usage compared to a fully decentralized process.