Grid'5000 user report for

Jump to: navigation, search

User information

( user)
More user information in the user management interface.

Experiments

  • Fractal Deployment Framework (Middleware) [in progress]
    Description: Fractal Deployment Framework (FDF) is a generic deployment framework allowing the automatization of the execution of heterogenous tasks composing the whole deployment process. Deployment can be defined as a set of tasks to realize such as software installation/uninstallation on remote nodes and activation/deactivation of software instances. These deployments tasks have to be orchestrated in a specific order: for instance, software activation cannot be done until it has been downloaded and configured. Thus, software deployment on large infrastructure as Grid5000, composed of thousands of nodes, is a complex task. Grids users can spend a lot of time (from hours to days) to prepare, reserve and configure nodes, install software, binaries or code, and finally start their experiences. FDF allows users/deployers to just describe their configuration instead of programming/scripting the deployment or taking care of dependencies or synchronization between deployment tasks. As a consequence, automatization of heterogeneous deployment activities becomes possible. FDF, implemented using the Fractal component model, is a “generic” deployment framework which means is independant of both: 1- The granularity of the software to deploy: middleware, applications, software components, objects, services, etc., and 2- The underlying technology, e.g., J2EE, CCM, Fractal, Web Services, OSGi, Object-Oriented applications, etc.
    Results: We have experimented our framework for the deployment of OpenCCM middleware, our free open source Java implementation of the OMG CORBA Component Model (CCM) specification, and CCM-based applications on many clusters of Grid5000. Deployment process of OpenCCM middleware on hundreds of Grid5000 nodes consists of many elementary tasks to orchestrate -accessing resources, reserving nodes and getting list of allocated nodes, -install OpenCCM application servers on nodes, plus all required libraries such as Java Runtime Environment and CORBA ORB, -configure for each node the execution environment by setting some specific shell variables, -start an application server (that will hosts business components) on each node, as a Java process, and useful middleware services such as a CORBA Naming Service, taking care of dependencies and synchronization problems, -instantiate application, -stop, -uninstall, -release resources, -etc. Using our framework, user simply describes his configuration in a simple natural language.
    illustrating chart picture not found

Publications

  • Deploying on the Grid with DeployWare [2008] (international)
    EntryType: inproceedings
    Author: Areski Flissi and Jeremy Dubus and Nicolas Dolet and Philippe Merle
    Booktitle: 8th International Symposium on Cluster Computing and the Grid (CCGRID'08)
    Month: May 18--22
    Address: Lyon, France
    Website: http://ccgrid08.ens-lyon.fr
    Note: To Appear
    Lastname: Flissi
  • A Generic Deployment Framework for Grid Computing and Distributed Applications [2006] (international)
    EntryType: inproceedings
    Author: Areski Flissi and Philippe Merle
    Booktitle: International Symposium on Grid computing, high-performAnce and Distributed Applications (GADA’06)
    Month: nov
    Address: Montpellier, France
    Longnotes: 20060614
    Website: www.cs.rmit.edu.au/fedconf/index.html?page=gada2006cfp

Collaborations

    Success stories and benefits from Grid'5000

    • Success stories
    • Deployment of OpenCCM middleware and CORBA Components-based applications up to near than 1000 nodes of Grid5000 have been successfully tested. For this case, effective deployment time (without the overhead introduced by the framework when parsing the user’s configuration file and that is due to the Fractal ADL tool) takes less than 300 seconds. Results are showed in the figure below. Grid5000 “virtual” nodes means here that many application servers may have been deployed on a same physical node.
    • Overall benefits
    • A large number of nodes.

    last update: 2008-03-26 17:21:11

    Personal tools
    Namespaces

    Variants
    Views
    Actions
    Public Portal
    Users Portal
    Admin portal
    Wiki special pages
    Toolbox