Adaptable decentralized orchestration engine for block structured non-transactional workflow in service oriented architecture.

In the Service Oriented Architecture (SOA), business processes are executed by nonscalable centralized orchestration engines. Nonetheless proliferation of business process applications in organizations raises scalability requirements. Decentralized orchestration engines are applied to address the scalability by decentralizing a process into design-time static fragments without considering runtime requirements. The fragments are then encapsulated into runtime components such as agents. The SOA orchestration layer suffers from the lack of adaptability with runtime environment in decentralization of business processes. Accordingly, three aspects of runtime adaptability in decentralization are studied in this thesis. The first aspect is frequent-path adaptability, which is equal to detecting closely-interrelated activities and encapsulating them in the same fragment. Another aspect is proportionalfragment adaptability, which is analogous to the proportionality of produced fragments with number of machines. The last aspect is available-bandwidth adaptability, which is process fragmentation based on current circumstances of communication media. An ever-changing runtime environment along with the mentioned adaptability aspects raises the following research problems: 1) there is no framework to support architectures, decentralization methods, and a feedback loop from runtime environment; 2) current decentralization methods do not consider the frequent-path and proportional-fragment adaptability aspects in creating fragments; 3) there is no algorithm to map runtime circumstances to a suitable decentralization method in order to satisfy the available-bandwidth adaptability. Accordingly, the following research objectives are considered: first, to propose a framework including architectures, decentralization methods, and a feedback loop from runtime environment; second, to improve response-time and throughput of decentralized business processes applying the frequent-path and proportional-fragment adaptability aspects; third, to improve bandwidth-usage of decentralized business processes applying the available-bandwidth adaptability. The contributions of this research are also as follows: i) An Adaptable and Decentralized Workflow Execution Framework (ADWEF) is introduced that proposes an abstraction of a runtime adaptable decentralization in the SOA orchestration layer; ii) two architectures Type-1 and Type-2 are presented for the ADWEF that are able to support the execution of dynamically created fragments; iii) three aspects of runtime adaptability in decentralization namely frequent-path, proportional-fragment and available-bandwidth are introduced; iv) two decentralization methods called Hierarchical Process Decentralization (HPD) and Hierarchical and Intelligent Process Decentralization (HIPD) are presented, which are capable of providing various fragments. The latter considers the frequent-path adaptability and both of them together satisfy both frequent-path and proportionalfragment adaptability aspects; v) A Fuzzy Decentralization Decision Making algorithm (FDDM) is presented based on the fuzzy logic to choose a suitable method of decentralization that satisfies the three adaptability aspects frequent-path,proportional- fragment and available-bandwidth; and, vi) an algorithm is introduced for wiring of dynamic fragments. Evaluations of the three adaptability aspects in the ADWEF demonstrate that the frequent-path adaptability greatly improves response-time, throughput, and bandwidth-usage of decentralized business processes. The proportional-fragment adaptability proves that number of fragments must be proportional to the number of workflow engines machines. The available-bandwidth adaptability which is realized by the FDDM algorithm unifies the mentioned adaptability aspects and reduces the number of exchanged messages compared to other methods.

Preview PDF FSKTM 2011 32.pdf Download (825kB) | Preview

Actions (login required)

View Item