Net-Centric Software & Systems Consortium Kick-off Meeting Feb. 26-27, 2009 Dynamic Service Composition with QoS Assurance Jing Dong Farokh Bastani I-Ling Yen UTD jdong@utdallas.edu UTD bastani@utdallas.edu UTD ilyen@utdallas.edu Net-Centric Software & Systems Consortium Kick-off Meeting Problem Description • Dynamic and reconfigurable service compositions • Frequently changing business requirements and conditions • Dynamic and volatile nature of system and Web environments • Non-stop runtime environments • Modeling service compositions • Complexity in large service composition processes • Lack of service abstraction for capturing the relations among services • QoS assured net-centric systems • System QoS and dependability requirements are very important • System needs to dynamically adapt Automated and dynamic service composition • Tool support service compositions satisfying QoS requirements 7/12/2016 Net-Centric Software & Systems Consortium 2 Net-Centric Software & Systems Consortium Kick-off Meeting Existing Solutions • Service composition • BPEL4WS – Only Functional, no semantics • Semantic Web services (OWL-S) – Lack the support for capturing the relations among services • Service discovery • Individual service selection based on QoS criteria – Focus on the selection of individual services • Service selection based on composite QoS behavior – Consider greatly simplified QoS property aggregation functions • Service composition analysis • Hardware failures for reliability analysis – No critical software faults • Simple summation of fixed latency for timing analysis – No consideration of the impact of newly added flows 7/12/2016 Net-Centric Software & Systems Consortium 3 Net-Centric Software & Systems Consortium Kick-off Meeting Our Solution • Service composition • Abstract service, service modeling and transformation • New planning technique: ProcedurePlan • A planner that can generate procedures • Much more scalable than all existing planners • Service discovery • Configurable services for QoS Tradeoffs • Service composition analysis • • • • 7/12/2016 Compositional QoS analysis for service composition Service-driven compositional reliability model and timing analysis Mutual trust evaluation protocol for trustworthiness Three-level compositional QoS analysis technique Net-Centric Software & Systems Consortium 4 Net-Centric Software & Systems Consortium Kick-off Meeting Our Solution (cont.) Extended UDDI Concrete Web Services Goal Specification GUI Web services QoS-capabilities Functional Goals User System QoS-Req OWL-S Abstract Service Profile Process Functional Composer Ontology Analyzer Composition Plan Service Composition Logic Extended Planner Extended OWL-S Ontology CQAD Framework Specification Parsing Parsers for various QoS specification languages 7/12/2016 Pattern Generator Service Patterns Decision Making Genetic Algorithm Particle Swarm Algorithm Grounding OWL-S Abstract Service Generator Compositional QoS Prediction … Instance Instance Pool QoS-Assured Composition Plan Reliability Timing Prediction Prediction Algorithms Algorithms Net-Centric Software & Systems Consortium … 5 Net-Centric Software & Systems Consortium Kick-off Meeting Experimental Plan • Current status • Pieces of techniques and tools • Planner, QoS analysis framework with decision process, etc. • Future plan • Build an integrated system synthesis framework • Develop a robust planner and a problem modeling framework • Develop QoS compositional analysis techniques and tools • Investigate partial planning techniques • When the goal cannot be reached with existing components/services • Investigate the power of the synthesis framework using real-world examples 7/12/2016 Net-Centric Software & Systems Consortium 6 Net-Centric Software & Systems Consortium Kick-off Meeting Industry Member Benefits • Share the techniques and tools which can • Dramatically reduce software development cost and time • Automate software customization process • Greatly enhanced capability in Web service composition • Expected from industrial partners • Provide real-world applications/scenarios • Used to validate the applicability of our solutions • Need close collaboration to understand and potentially modify the case and the scenarios • Provide development time/cost estimates • Used to compare with the time/cost of our approach 7/12/2016 Net-Centric Software & Systems Consortium 7 Net-Centric Software & Systems Consortium Kick-off Meeting Deliverables and Budget Year 1 6 months Abstract service modeling. Defining service relationships. Dynamic updating service ontology and super instance. Algorithm for populating the instance pools. Categorizing the service parameters 6 months Pattern-based planning techniques for service composition. Pattern-based service composition. Service pattern extraction. Planning techniques for complex compositions. Year 2 6 months QoS-assured service composition. Service-centric compositional real-time analysis. Service-driven compositional reliability analysis. Trustworthiness analysis for service composition. Configurable web services model. 6 months Integrate the planner, synthesis environment, and QoS analysis tools. Perform development cost/time analysis and comparison. Case studies to validate the approach. (Requires close collaboration with industrial partners to provide the case study, discuss the scenarios and potential modifications, provide cost/time data of the existing development process.) Duration Total 1 Graduate Studentsat $25000 stipend + $10000 Tution per year 2 years 70000 0 Faculty months at $10000 per faculty effort 2 years Travel at 1500 per conference travel 2 years 3000 Total Cost 73000 7/12/2016 Net-Centric Software & Systems Consortium 8 Abstraction Modeling Analysis Semantic Web Services Dynamic Service Composition Service Selection with QoS requirements Analysis with QoS Assurance NEW INSIGHTS vote Abstract service captures the relationships among the services and raise the level of abstraction Service selections should consider not only the functional and semantic aspects, but also QoS requirements Service composition analysis with QoS assurance Techniques and tools to analyze and evaluate these compositions MAIN ACHIEVEMENT: Developed abstract service to allow service composition and analysis at a higher level Developed compositional genetic algorithms to efficiently select compositions that can optimally achieve QoS requirements Developed techniques to evaluate and analyze service composition with QoS assurance HOW IT WORKS: QUANTITATIVE IMPACT Service composition The most challenging issues in reconfigurable service composition with QoS assurance: (1) How to provide a holistic view of multiple QoS properties under changing business requirements? (2) How to dynamically compose the web services that satisfy QoS criteria? (3) How to analyze the service compositions? Our techniques provide partial answers to these issues. ASSUMPTIONS AND LIMITATIONS: • Each service has well-defined interface that describes the functional and semantic properties • Service implementation is hidden from its interface • Generalization of our approach to many applications is to be validated Comparing results computed by different composition algorithms on different QoS criteria • Develop a tool kit to allow automated composition of web services • Develop evaluation suite to evaluate and analyze service composition with QoS requirements • Validate our approach on many real-world applications END-OF-PHASE GOAL STATUS QUO Net-Centric Software & Systemsdescription Consortium Topic/project/effort Kick-off Meeting A tool kit for composition of web services with QoS assurance • Modeling service composition at a higher level of abstraction • Developing algorithms for service selections with QoS criteria • Analyzing service composition • Evaluation techniques can effectively assess the service composition with a high degree of confidence, and can be used to guide the system design process Automated service composition can significantly save development time and improve system quality