Font Size: a A A

Research On Planning Based Web Services Composition And Related Technology

Posted on:2010-09-17Degree:DoctorType:Dissertation
Country:ChinaCandidate:H YangFull Text:PDF
GTID:1118360278965462Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
Service oriented computing (SOC) is rapidly becoming the prominent paradigm for distributed computing and electronic business applications. SOC allows for service providers and service application developers to construct value-added services by service composition, which combines existing services that are resident on the Web. For automatic service composition, we research on several critical questions based on constraint satisfaction.Supported by National Nature Found of "Mobile Intelligent Service Platform" , National Basic Research Priorities Program (973) of "The Framework of Authentic and Integrative Web Service", and Beijing Education Department Project "PingGu Traveling Service Platform", we begin service composition research with two major techniques, service reasoning and constraint satisfactions. A Planning based model is provided, extended POPLAN algorithm is analyzed as its resoiver, and a three layered framework is presented for its implementation. Testing cases in PingGu Traveling Service Platform show that service composition can be achieved automatic or semi-automatic based on the algorithm and framework. The main contributions are as follows:(1) For automatic service composition based on resolving service reasoning and constraint satisfaction, a STRIPS based planning model is provided, and Extended POPLAN is provided as its resolver. The novelty of this algorithm includes: (a) service model is described as a NFA (non-determined finite automata), which determined part is analyzed by Casual-link path searching between initial node and end node. (b) Non-determined part of service model is divided into non-determined request and non-determined response. With non-determined request treated by establishing precondition and non-determined response by abstract BPEL/UDDI interaction, non-determined finite automata is transformed into determined finite automata, and can be treated like step (a).(2)Based on Extended POPLAN for automatic service composition, a three layered service composition framework (MOCIS) is provided. Service reasoning is resolved by Activity Tree Planner, describing service activities and their structures. Constraint satisfactions are divided into quality constraint satisfactions and quantity constraint satisfactions. Message Planner resolves quality constraints by interaction UDDI for tModel information, determining service messages and their correlation set. After Activity Tree Planner and Message Planner execution, abstract BPEL list is achieved. Concrete BPEL list is created within the third layer Scheduler. Scheduler resolves quantity constraint satisfaction based on a backtrack algorithm named as ChooseProvider, for choosing accepted service providers for each activity and related role. With activity, structure, message, correlation set and service provider determined, concrete BPEL list is created. User can choose one of them, deploy it into application server and get the composite service response. This is a semi-automatic framework for three layers are all automatic, with two user interaction, abstract BPEL selection and concrete BPEL selection.(3)In order to composite service according to complex business logic, a spatial-temporal object model and service clustering composition approaches is proposed. Different from traditional traveling service composition taking less account of "WHILE" structure, this approach is a several days' composition, which is a "WHILE" structured BPEL and with two constraints for scenery spots in a day are as near as possible, and should be far and different for different days. Inspired by clustering characteristics, objects inside a cluster have much similarity and in different clusters have much difference, a K-means based service clustering is provided as a resolver for multi-days service orchestration. Testing cases show that other than providing "What" and "How" capabilities with traditional methods, service clustering can resolve "When" and "Where" capabilities for service orchestration.Different testing cases within this thesis are deployed and run within "Pinggu Traveling Service Platform", including Extended POPLAN testing, three layered architecture testing, service clustering testing, and so on.
Keywords/Search Tags:service reasoning, service constraint satisfaction, planning based service composition, non-determined service composition, extended POPLAN, three layered service composition framework, abstract BPEL, service clustering
PDF Full Text Request
Related items