BP ICIW07

QoS Composition of Services for Data-Intensive Application : QoS Composition of Services for Data-Intensive Application Antonio Bucchiarone PhD Student – IMT Graduate School Piazza S. Ponziano 6, 55100 Lucca (Italy) antonio.bucchiarone@imtlucca.it and Luigi Presti Software Engineer IBM Software Group - Tivoli Rome Lab Via Sciangai 53, 00144 Rome (Italy) luigi.presti@it.ibm.com


Agenda : Agenda Introduction Background QoS Model for Service Composition A Front-End Application Queueing Model Service Composition Flow Models QoS Service Composition Algorithm Related Work Conclusions and Future Work


Introduction : Introduction SOC is a promising means to integrate heterogeneous systems Services from different providers can be integrated into a composite service QoS of Data-Intensive applications Reliability Performance Service description QoS dynamic composition


Background - I : Background - I Services (functionality) Context adaptive and intelligent user services Information services Intermediary services Location-based services Services (technical protocols) Web Services Grid Services “a service provides some useful functionality through a well defined interface and it is possible to combine them to produce useful “composite” services.”


Background - II : Background - II Services Composition Web Services SOAP, WSDL and UDDI Orchestration (BPEL4WS) and Choreography (WS-CDL) Grid Services To utilize the power of heterogeneous distribute resources, computing resources, data storage systems, instruments, .. Grid and Web Services are converging in the WSRF A series of specifications for performing grid computing on top of web services


Background - III : Background - III Service Level Agreement (SLA) It complements a service description language It defines the agreed performance characteristics and the way to evaluate and measure them


QoS Model for Service Composition : QoS Model for Service Composition Front-End application The actors and theirs main functionalities Queueing Model For data-intensive applications Services Composition Flow Models Types of services and basic relationships QoS parameters QoS Service Composition Algorithm To compose services that have QoS attributes


QoS Model for Service Composition Front-End application - I : QoS Model for Service Composition Front-End application - I “A system that is able to put together different services from various providers in order to construct and execute a data-intensive application” Providers Companies distributed on the net Customer Client that wants execute a new application that satisfies some QoS characteristics Front-End application System that realizes the application


QoS Model for Service Composition Front-End application - II : QoS Model for Service Composition Front-End application - II Service Description (providers) QoS parameters for the service classification Service Classification (front-end) It subdivides the services in classes based on their description and QoS attributes Workflow definition (front-end) Services composition structure Meta-workflow (data-flow, Pipeline) Application Chosen (Customer) It choose the kind of application that he wants (data mining, astronomy, traveling, tourism,..) QoS Inizialization (Customer) max-cost of the service composition min-perf : medium departure time of each result in the last node of the workflow QoS Algorithm (front-end) QoS Services Composition Input: Workflow definition and QoS parameters Output: workflow in which the services are instanciated (final application) Searching Services (front-end) Search a service based on QoS parameters Execution Throw Exception


QoS Model for Service Composition Queueing Model : QoS Model for Service Composition Queueing Model From: L. Kleinrock, “Queueing Systems, Vol. I: Theory,” The arrival process of customers (Ta) The service times (Ts) The service discipline Fifo Lifo Random Processor sharing The queue occupation rate or server utilization ρ = Ts / Ta Our model: Data-Flow Our target: ρ < 1


QoS Model for Service Composition Services Composition Flow Models - I : QoS Model for Service Composition Services Composition Flow Models - I Types of Services Relationships (DAG) Data-flow Or Pipeline


QoS Model for Service Composition Services Composition Flow Models - II : QoS Model for Service Composition Services Composition Flow Models - II Goal: build an application from an automatic composition of services QoS parameters Intuitive Easy to measure


QoS Model for Service Composition QoS Service Composition Algorithm : QoS Model for Service Composition QoS Service Composition Algorithm Input Workflow definition of the service composition, and QoS attributes Output Workflow instantiated with real services satisfying QoS attributes


Related Work : Related Work SWORD project Rule-based Web Services composition Offline composition (not at run time) FUSION framework Web Services composition according users satisfaction criteria Run time composition WebQ framework Adaptive management of Web services QoS selection criterion only considers service load and makes only local decisions EFlow A platform for the specification, enactment and management of composite services (graphs) The graphs may include service, decision and event nodes GSFL (Grid Service Flow Language) Grid Services composition


Conclusions : Conclusions A front-end application in order to develop data-intensive applications starting from services developed by different owners QoS composition algorithm Open Points To enrich composition model with more complex workflows To enrich QoS composition algorithm considering the transmission time parameter that can be variable To implement the front-end application with the QoS composition algorithm in order to use it in a real case-study


Slide16 : Thank you for your attention! Antonio Bucchiarone PhD Student – IMT Graduate School Piazza S. Ponziano 6, 55100 Lucca (Italy) antonio.bucchiarone@imtlucca.it and Luigi Presti IBM Software Group – Tivoli Rome Lab Via Sciangai 53, 00144 Rome (Italy) Luigi.presti@it.ibm.com