Research On Generic Web Service Composition And Selection Based On Contracts

With the development of SOA (service-oriented architecture), Web service composition provides the features of flexibility and scalability to meet the complex requirements. When services are combined, more attention is paid to regulate the business interactions between service suppliers. In this thesis, transactions are regulated by binding contracts specifying ob-ligations of the parties.There are three key challenges about contract-regulated Web service composition. The first challenge is how to seamlessly integrate behaviour descriptions with contracts. The second challenge is how to ensure contract-regulated reliable execution without repeating the proof. The third challenge is how to select the best instantiation of contract-regulated service composition with accuracy and high speed. Referring to above challenges, this thesis proposes a novel generic model to integrate behaviour descriptions with contracts and batch proof me-thod to improve the reliability for service compositon and a creative D-GSS (discrete group search service) model for high-effective service selection.The main contributions of this thesis can be described as follows:(1) Generic service model describing contract regulated Web service composition is created to reduce the need for proof. Generic service model based on tecton expressed using "concept descriptions" applies techniques with concepts to capture general properties of ordi-nary service composition and abstract the contracts governing them as generic forms. Verified generic specification at an abstract level can be used in a variety of instances without repeat-ing the proof, and there will be a plenary promotion of reusability.(2) We propose the definition of true situation and batch proof method used to verify completeness of generic specification’s properties with more efficiency. We propose the defi-nition of true situation based on contract regulated conditions and design batch proof method for verifying completeness of generic specification’s properties in all true situations at the same time to ensure service composition reliability throughout the life cycle. This thesis pro-vides an effective method to take into account all the situations the composition may expe-rience simultaneously. If all running paths succeed, the specification is complete. Otherwise the specification must be extended. With the help of Violet system, it is easy for the user to find and understand proof procedure of batch proof method. A library is built to keep verified generic specifications for the purpose of instantiation and reusability.(3) Cost function is difined to precisely estimate the quality of every instantiation of contract-regulated service composition while multiple QoS constraints must be satisfied. In D-GSS model, to satisfy the restricted demands for QoS attributes of contract-regulated ser-vice composition, cost function is defined to easily calculate the value of every instantiation’s quality. The better the instantiation is, the lower the value of the cost function is. The follow-ing steps are involved in defining the cost function:each QoS attribute must be quantitative; every global QoS attribute is aggregated by the QoS attributes of all atomic services consi-dering the integration relationships of service composition; the utility function synthesizes the cost and benefit criteria of QoS attributes and describes the relationship between global QoS attributes and corresponding constraints; at last the cost function is based on the values of the utility function and the weights the user defined.(4) DC-GSO (discrete group search optimizer) algorithm will benefit searching for the best instantiation of contract regulated service composition quickly. In D-GSS model, DC-GSO algorithm inspired by animal searching behavior is specially created to solve dis-crete optimization problems for service selection. This algorithm is applied to rapidly choose the best instantiation for the service requester from a large number of instantiations of generic service composition. The group members in this algorithm are empolyed as producer, scrounger or ranger to search for the global optimum with the lowest value of cost function after many iterations. About optimal searching strategies for contract regulated service selec-tion, DC-GSO algorithm has advantage in competition to other algorithms in terms of accu-racy and convergence speed.