Data-Aware Service Choreographies through Transparent Data Exchange

Transcription

1 Data-Aware Service Choreographies through Transparent Data Exchange University of Stu/gart Universitätsstr Stu/gart Germany Michael Hahn, Dimka Karastoyanova and Frank Leymann InsQtute of Architecture of ApplicaQon Systems Phone Fax

2 Mo%va%on Simulation lattice energy Prepare Input Files opal_in Run Simulation snapshots saturation Process Snapshot Video Plot video plot Prep MC Snapshot Processing snapshots[i] Search Atom Clusters cluster[i] Determine Position and Size possize[i] Visual CLUS XYZR Service Task Data Object Data Object Collection Parallel Gateway Message Start Event End Event Message End Event [1] Binkele and Schmauder: An atomistic Monte Carlo simulation for precipitation in a binary system. In: International Journal for Materials, 94, pp. 1-6, [2] Sonntag Michael et Hahn al.: Using Services and Service Compositions to Enable the Distributed Execution of Legacy Simulation Applications. In: Proc. of ServiceWave'11. 2

3 Mo%va%on t i t k t Simulation lattice energy Prepare Input Files opal_in Run Simulation snapshots saturation Process Snapshot Video Plot video plot Prep MC Snapshot Processing snapshots[i] Search Atom Clusters cluster[i] Determine Position and Size possize[i] Visual 1 Anticipatory data transfer CLUS XYZR Service Task Data Object Data Object Collection Parallel Gateway Message Start Event End Event Message End Event [1] Binkele and Schmauder: An atomistic Monte Carlo simulation for precipitation in a binary system. In: International Journal for Materials, 94, pp. 1-6, [2] Sonntag Michael et Hahn al.: Using Services and Service Compositions to Enable the Distributed Execution of Legacy Simulation Applications. In: Proc. of ServiceWave'11. 3

4 Mo%va%on t i t k t Simulation lattice energy Prepare Input Files opal_in Run Simulation snapshots saturation Process Snapshot Video Plot video plot Prep 2 MC Avoid unnecessary routing Snapshot Processing snapshots[i] CLUS Search Atom Clusters cluster[i] XYZR Determine Position and Size possize[i] Visual 1 Anticipatory data transfer Service Task Data Object Data Object Collection Parallel Gateway Message Start Event End Event Message End Event [1] Binkele and Schmauder: An atomistic Monte Carlo simulation for precipitation in a binary system. In: International Journal for Materials, 94, pp. 1-6, [2] Sonntag Michael et Hahn al.: Using Services and Service Compositions to Enable the Distributed Execution of Legacy Simulation Applications. In: Proc. of ServiceWave'11. 4

5 Mo%va%on 3 Multicasting t i t k t Simulation lattice energy Prepare Input Files opal_in Run Simulation snapshots saturation Process Snapshot Video Plot video plot Prep 2 MC Avoid unnecessary routing Snapshot Processing snapshots[i] CLUS Search Atom Clusters cluster[i] XYZR Determine Position and Size possize[i] Visual 1 Anticipatory data transfer Service Task Data Object Data Object Collection Parallel Gateway Message Start Event End Event Message End Event [1] Binkele and Schmauder: An atomistic Monte Carlo simulation for precipitation in a binary system. In: International Journal for Materials, 94, pp. 1-6, [2] Sonntag Michael et Hahn al.: Using Services and Service Compositions to Enable the Distributed Execution of Legacy Simulation Applications. In: Proc. of ServiceWave'11. 5

6 Outline n Data-Aware Choreography Management Life Cycle n Architecture n Challenges n Conclusions and Outlook Michael Hahn 6

7 Data-Aware Choreography Management Life Cycle Analysis Modeling CM CDM Monitoring Transparent Data Exchange Refinement Abstract Wf Model CDDG Execution Deployment Executable Wf Model PDDG Deployment Bundle Enhanced Phase Methods Software Artifact Michael Hahn 7

8 Introducing a Choreography Data Model (CDM) Choreography Data Model lattice energy opal_in saturation snapshot cluster possize plot video Simulation Prep #energy lattice energy #opal_in Prepare Input Files MC opal_in #lattice, #opal_in #snapshots[], #saturation Run Simulation Snapshot Processing snapshots[i] snapshots saturation #snapshots[i] Search Atom Clusters Process Snapshot cluster[i] #cluster[i], #possize[i] Determine Position and Size possize[i] Video #video #snapshots[] Plot #plot #saturation CLUS XYZR Visual video plot #snapshots[i] #cluster[i] #cluster[i] #possize[i] Service Data Object Data Object Parallel Message End Event Message TraDE Michael Hahn Task Collection Gateway Start Event End Event Data Object 8

9 Data-Aware Choreography Management Life Cycle Analysis Modeling CM CDM Monitoring Transparent Data Exchange Refinement Abstract Wf Model CDDG Execution Deployment Executable Wf Model PDDG Deployment Bundle Enhanced Phase Methods Software Artifact Michael Hahn 9

10 Data Dependencies between Par%cipants (CDDG) MC Prep lattice snapshots video energy Prepare Input Files opal_in Run Simulation Simulation Process Snapshot Video Plot plot Visual saturation snapshots[i] snapshots[i] Snapshot Processing Search Atom Clusters Determine Position and Size cluster[i] possize[i] possize[i] CLUS XYZR cluster[i] cluster possize Michael Hahn 10 cluster[i]

11 Data-Aware Choreography Management Life Cycle Analysis Modeling CM CDM Monitoring Transparent Data Exchange Refinement Abstract Wf Model CDDG Execution Deployment Executable Wf Model PDDG Deployment Bundle Enhanced Phase Methods Software Artifact Michael Hahn 11

12 Par%cipant Data Dependence Graph (PDDG) n Analysis of internal data flow and data dependencies regarding the defined data objects in refined workflows n Knowledge can be used as input for opqmizaqon during modeling Qme and run Qme n PDDGs enable refinement of CDDG Michael Hahn 12

13 Refined Choreography Data Dependence Graph (CDDG) MC Prep lagce snapshots video energy Prepare Input Files opal _ in Run Simula%on size(snapshots) SimulaQon Process Snapshot Video Plot plot Visual saturaqon snapshots [ i ] snapshots [ i ] Snapshot Processing Search Atom Clusters Determine Posi%on and Size cluster [ i ] possize [ i ] possize [ i ] CLUS XYZR cluster [ i ] cluster possize Michael Hahn cluster [ i ] 13

14 Architecture PDDGA TraDEA TraDEZ TraDEY TraDEX TraDEB PDDGB WfMA WfMSA Execution & Monitoring Deployment WfMSB WfMB Choreography & Orchestration Modeling Environment A Modeling & Refinement Choreography & Orchestration Modeling Environment B WfM Workflow Model PDDG Participant Data Dependence Graph WfMS Workflow Management System Data Exchange Message Exchange Michael Hahn 14

15 Challenges Data analysis and optimization algorithms based on introduced new artifacts, i.e., CDM, CDDG & PDDG Incorporating new knowledge into model transformation process Concept and implementation of a distributed TraDE middleware Communication protocols for fault-tolerant, secure and reliable asynchronous data exchange Integration with SOA middleware (ESB, WfMS) and services Data correlation and monitoring Definition of optimization strategies for, e.g., data staging, placement and exchange during run time Michael Hahn 15

16 Conclusions and Outlook n Data-aware service choreography management life cycle with TraDE methods for data flow analysis and opqmizaqon n System architecture that implements the life cycle n challenges represent our road map for future work Michael Hahn michael.hahn@iaas.uni-stuttgart.de Michael Hahn 16