XtreemOS summit 2010

Program

Our second XtreemOS summit will be one day in length, on August 30, and organized as follows:

  • In the morning:
    • 9:30-9:35: summit opening (Christine Morin)
    • 9:35-11:00: tutorial "Easing Application Execution in Grids with XtreemOS Operating System" (part 1/2) - see below for details
    • 11:00-11:30: coffee break
    • 11:30-13:00: tutorial (part 2/2)
  • 13:00-14:30: Lunch
  • In the afternoon:
    • 14:30-15:40:  XtreemOS demos (demonstration of XtreemOS functionalities, demonstration of real applications (scientific applications)) - Yvon Jégou (INRIA), Barry MacLarnon (SAP)
    • 15:40 - 16:00: open testbed presentation (Yvon Jégou - INRIA)
    • 16:00-16:30: break
    • 16:30 - 17:15: presentations of experimental results: computing challenge using XtreemOS (webpage).

- Alvaro Parra, Exequiel Sepulveda & Felipe Lema (ALGES lab, Universidad de Chile)

- Syed Raheel Hassan, Jasmina Pazardzievska, Maxime Syrame and our supervisor Prof.Julien Bourgeois from LIFC (Université de Franche-Comté, France).

- Emanuele Carlini, Sebnem Erturk & Giacomo Righetti (University of Pisa, Italy)

    • 17:15 - 17:45: open discussions
    • 17:45: closing remarks

XtreemOS tutorial

This tutorial presents XtreemOS Grid system from the user point of view. Its goal is to show why XtreemOS is a system of choice for Grid users who want to easily execute their applications while taking advantage of advanced features for job management. We show what can be done with XtreemOS with regards to application execution, how it can be done and how the available features are implemented. Demonstrations will be presented throughout the tutorial.

XtreemOS is a Linux-based operating system that includes Grid functionalities. It is characterised by properties such as transparency, hiding the complexity in the underlying distributed infrastructure; scalability, supporting hundreds of thousands of nodes and millions of users; and dependability, providing reliability, highly availability and security. XtreemOS system is available for clusters, standalone PC and mobile devices.

XtreemOS implements a job-oriented scheduling within the subset of resources obtained by the resource discovery mechanism. To ease the use of the Grid services, XtreemOS mimics the well-known Linux functionality as opposed to offering different abstractions and functionality, which are more oriented to the Grid. For instance, AEM implements job control through signals. The reservation service implemented by AEM provides resource co-allocation allowing the execution of distributed/parallel applications on multiple Grid nodes. Reservations can be dynamically modified. AEM also provides flexible and accurate job monitoring. It manages job dependencies, providing an interface to external workflow engines. Lastly, XtreemOS fully supports the execution of interactive
applications. The XtreemGCP Grid checkpointing service takes care of reliable execution of distributed applications that can take advantage of rollback recovery protocols. XtreemGCP can checkpoint/restart and migrate applications running on multiple heterogeneous Grid nodes. Various kernel checkpointers are supported through a library implementing a common checkpointer interface.  Being an operating system XtreemOS is able to execute any kind of application in the business and scientific domains including workflows, distributed applications, parallel applications such as MPI applications and SAGA applications.

Target audience: Grid users, Grid application developers, Grid system and middleware designers

Duration: 90'x2

Tutorial presenters: Christine Morin (INRIA), Thilo Kielmann (VUA), Yvon Jégou (INRIA), Jacobo Giralt (BSC)

  • Tutorial programme

- Overview of XtreemOS
- Dynamic resource discovery for job execution
- Resource management
        - Resource selection
        - Resource reservation
        - Resource (co-)allocation
- Job Management
    - Job control
    - Job dependency management
    - Job monitoring & auditing
    - Support for interactive applications
    - Examples of job execution
        - Workflow
        - MPI applications
- Reliable job execution
    - XtreemOS-GCP checkpointing service for distributed applications
        - Integration of different checkpointer packages
        - Generic callbacks for application-level optimizations
        - Channel checkpointing with heterogenous checkpointer packages
- Executing SAGA applications on XtreemOS

XtreemOS demos

The demo will highlight technical functionalities of the XtreemOS system and present real applications (scientific applications):

  • Using XtreemOS’s VOLifeCycle web application to create and to operate a Virtual Organization,
  • Submitting jobs in XtreemOS,
  • Using XtreemFS, XtreemOS file system.

 

The main benefits that this demo aims to bring are:

  • discovering the advantages of introducing Grid and cloud functionalities at the operating system level,
  • getting insight on Grid security issues and learning how to create and operate virtual orga- nizations in a simple way in XtreemOS,
  • finding out how to run complex distributed applications with the aid of the XtreemOS Application Execution Manager,
  • learning how to use the distributed file system XtreemFS for storing data in a reliable manner.

 

The main functionalities of XtreemOS will be demonstrated using virtual machines running on a few laptops connected through a local ethernet switch. Whenever possible, we will also demonstrate this operating system running on a bunch of servers located in XtreemOS partners premises and distributed across Europe.