Mathematical theory of service composition and service networks
Open Access
- Graduate Program:
- Industrial Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- December 15, 2010
- Committee Members:
- Soundar Kumara, Dissertation Advisor/Co-Advisor
- Soundar Kumara, Committee Chair/Co-Chair
- Arunachalam Ravindran, Committee Member
- Tao Yao, Committee Member
- Hong Xu, Committee Member
- Reka Z Albert, Committee Member
- Dongwon Lee, Committee Member
- Keywords:
- dynamic programming
- service composition
- complex networks
- mathematical theory
- concept service (CS) network matrix
- goal programming
- enterprise integration
- Abstract:
- This dissertation addresses the theoretical foundations of service composition and develops network based techniques, mathematical programming, and stochastic models for service composition. For the first time, a mathematical theory, which forms the foundation to define service science as a rigorous discipline is explored. The theorems of solution existence and convergence are generated. Mutuality/Duality theory of service space and concept space is established. The Concept Service (CS) network matrix is developed to study the network structure of service composition. The potential of CS network provides insights to identify important services and concepts. This information is used to connect the initial query with sub-components in CS network, thus reducing the computational time for the planning algorithms. A set of heuristic service composition algorithms is developed based on CS network matrix. Based on the mathematical theory established, three types of multi-criteria programming models are designed to find optimal service composition solutions for the offline-planning problem. The three types of mathematical programming models are: Multi Criteria Programming(MCP), Multi Criteria Goal Programming for Optimal composition(MCGPO) and Multi Criteria Goal Programming for Non-optimal composition(MCGPN). MCP model can generate an optimal solution if solutions satisfying all the customer’s functional and nonfunctional requirements exist. In addition, the MCGPO model allows to automatically select a trade-off among the objectives according to the customers’ preferences. MCGPN model is fast for generating satisficing solutions. Stochastic models are developed to compose service queries under uncertainty. In these models, nested compound work flow and service workload are considered. The algorithms directly utilize prior information of the potential of Concept Service (CS) network matrix to generate a transition matrix. This offers a fast and convenient approach for real time service composition to capture the uncertainty and dynamic phenomena. Finally, possible applications of service composition in product design, global manufacturing, supply chain and resource allocation are discussed.