Presentation on theme: "Enhanced Middleware Architecture for the Remote Instrumentation e-Infrastructure OGF28 Alexey Cheptsov High Performace Computing Center."— Presentation transcript:
Enhanced Middleware Architecture for the Remote Instrumentation e-Infrastructure OGF28 Alexey Cheptsov High Performace Computing Center Stuttgart (HLRS), Germany on behalf of the DORII Project March 17, 2010, Munich
Middleware Architecture for Remote Instrumentation (c) BMW e-Infrastructure for Remote Instrumentation
Middleware Architecture for Remote Instrumentation OUTLINE Introduction Requirements to the middleware DORII middleware architecture Main components of the architecture grid services for remote instrumentation e-Infrastructure front-ends application development environment workflow management system parallel application support solutions for scientific visualization User and developer support
Middleware Architecture for Remote Instrumentation Application requirements to e-Infrastructures middleware A full-featured GUI toolkit that enables easy access and discovering of grid resources and services, and serves a rich user collaborative suite A workflow management and monitoring system for complex scientific experiments An integrated application development environment, with enhanced debugging and deployment features. A visualization client for on-grid data analysis Parallel application implementation libraries, with integrated performance analysis possibilities and back-ends for visualization of performance characteristics.
Visualization and Interactivity (Int.EU.Grid) Workflow Management (VLab) Grid Services Core middleware stack (EGEE) Computation Grid Storage Grid Application Development on Grid (g-Eclipse) Grid access and collaboration (GRIDCC) DORII approach Remote Instrumentation on Grid (GRIDCC) DORII Middleware Architecture User-level middleware tools Middleware Architecture for Remote Instrumentation Parallel application support (Open MPI)
Middleware Architecture for Remote Instrumentation DORII Middleware Architecture Main features IE extends core middleware by interconnection of instruments with the infrastructure Common Library collects together classes and interfaces for obtaining access to the underlying core services (+ IE) VCR act as a central user front-end to the e-Infrastructure, facilitating resource browsing, job management etc. WfMS enables design and management of the application workflow Visualization and interactivity providing components greatly extend application support within the e-Infrastructure Open MPI and MPI-Start support parallel MPI applications
Middleware Architecture for Remote Instrumentation Remote Instrumentation Services on the Grid The middleware to interface a sensor/instrument INSTRUMENT ELEMENT (IE) Represents a virtualization of diverse data sources and provides the traditional Grid with an abstraction of a real instrument/sensor Provides user clients with an interactive interface to remotely control an instrument.
Middleware Architecture for Remote Instrumentation Access to grid services Common Library offers simple client-side API for accessing core middleware + IE services deployed on the infrastructure equipped with a single access point for the user-side middleware – a CL client simplified installation procedure for both client and user sides utilized by client-side middleware components currently compatible with gLite
Middleware Architecture for Remote Instrumentation e-Infrastructures GUI toolkit a full featured front end to the e-infrastructure and underlying services (IE, …) ready-to-use collaborative environment, offering a set of groupware tools (logbook, chat, wiki) entry point for further components (workflows, visualization etc.) VCR
Middleware Architecture for Remote Instrumentation Workflows for complex experiments WfMS supports users in defining, managing and monitoring remote instrumentation scenarios client-server architecture with light-weight client side
Middleware Architecture for Remote Instrumentation Workflow Management System (WfMS) WEB SERVICESWEB SERVICES User friendly Swing GUI Creating and editing workflow One-click submission on Grid Monitoring workflow execution Java and Web Services Manages workflow's jobs Common Lib used for Grid interaction Three layers design (manager, data base, Common Lib) server - Workflow Manager client - Workflow Editor and Monitor
Middleware Architecture for Remote Instrumentation e-Infrastructure Storage Resources User area Constant data User Frontend/VCR Remote Instrument Experiment properties Data from CINECA Computing resources INGV / CINECA INGV / CINECA IM 1 2 OPATM-BFM GVid 3 A workflow example
Middleware Architecture for Remote Instrumentation The WfMS architecture Three layers: workflow management database Common Lib (OSGi services) Comunication editor workflow management layer and workflow management layer OSGi services using web services Database connected with workflow management layer using Hibernate.
Middleware Architecture for Remote Instrumentation Workflow Editor Create the workflow VO-driven access to applications predefined default applications scenarios Grid browsing (SE, CE, IE) in workflow's data blocks Run the workflow on the Grid one-click workflow submission workflow's logic is managed by WfMS Job's output download Monitor execution progress presentation of state of each workflow's job
Middleware Architecture for Remote Instrumentation Workflow Manager Workflow execution steering Reading and traversing of workflow graph Translating workflow's jobs to JSDLs, IE and data operations (SE) Authorisation & Authentication operations on Grid using user's Proxy certificate MyProxy delegations to user's proxy accuired from VCR Users session persistence Operations on Grid Job submission, operations on IE, job status check, data operations done via Grid access libraries (gLite: g-Eclipse) SOA and Web Services
Middleware Architecture for Remote Instrumentation Integration with VCR 1.User launches Workflow Editor from VCR 2.VCR prepares information 3.User works with WF-E 1.WF-M identifies user with JNLP file and proxy delegation 2.WF-M monitors if proxy is valid and renews it when needed (MyProxy)
Middleware Architecture for Remote Instrumentation gLite Job management (JSDL/JDL, parametric JSDL) Remote file systems (gsiFTP, SRM, LFC) VOMS proxy extension, MyProxy support Application deployment Glue information system Instrument Elements support GRIA (industry Grid middleware): job management, GRIA file store Globus GT2: job management (RSL), gsiFTP, gLogin Amazon Computing Cloud: S3 storage grid middleware supported in g-Eclipse
Middleware Architecture for Remote Instrumentation g-Eclipse in DORII access and development platform that will be used whenever the application developers need an IDE for their sequential or parallel code an integrated and user-friendly middleware independent environment extends the native Eclipse IDE for different Grid actors (users, operators, developers) provides building blocks for construction of Common Library
Middleware Architecture for Remote Instrumentation Solutions for scientific data visualization A practical use case: OPATM-BFM simulation application the size of the produced short-term forecast data: 72 GB A practical use case: OPATM-BFM simulation application the size of the produced short-term forecast data: 72 GB
Middleware Architecture for Remote Instrumentation glogin enables online communication between nodes on the Grid and off the Grid provides shell functionality for accessing Grid nodes is a standard Grid job supports GSS-based encryption provides redirection of stdin/stdout/stderr, supports traffic forwarding GVid Visualization is rendered on the remote node and the output is streamed as MPEG-4 data to the client For interactivity the user interaction is transferred back to the render node Data transferred using glogin Works transparently with OpenGL/GLUT applications, provides VTK binding
Middleware Architecture for Remote Instrumentation GVid based visualization applet Implemented converter from NetCDF to VTK data format VTK plug-in is developed for g-Eclipse IDE
Middleware Architecture for Remote Instrumentation Parallel applications support A practical use case: OPATM-BFM simulation application Simulation duration for a standard use case: ~8 hours (64 CPU) A practical use case: OPATM-BFM simulation application Simulation duration for a standard use case: ~8 hours (64 CPU) The data collected from the instrument are post processed by simulation applications
Middleware Architecture for Remote Instrumentation Parallel applications support Open MPI PACX-MPI MPI Start Production quality middleware Full MPI-2 standard conformant Portable and maintainable Thread safe Network and process fault tolerant Enables running parallel applications on a metacomputer consisting of different CEs Installed on top of the main MPI library No changes in the application source code are required Improve starting parallel applications on the Grid Supports different MPI libraries, file systems, resource management and process launch systems Easy usable and extendable
Middleware Architecture for Remote Instrumentation Open MPI improved MPI-IO functionality support of PNetCDF library integration with performance analysis tools (Vampir, Valgrind) enhanced profiling possibilities thanks to PERUSE interface Message P1P2P32 Collectionofinput netCDFfiles Collectionofoutput netCDFfiles … I/Ooperations -passing communication
Middleware Architecture for Remote Instrumentation DORII Releases. Source code management system (subversion) File release system Bug trackers
Middleware Architecture for Remote Instrumentation User and developer support user support mailing list user support mailing list DORII middleware page DORII middleware page Bug and issue trackers
Middleware Architecture for Remote Instrumentation Questions Thank you for attention!