1 / 15

Distributing a n-body Problem Algorithm at Large-Scale over a Multi-Sites Grid using JavaSpace

Cracow Grid Workshop. October, 18th 2006. Distributing a n-body Problem Algorithm at Large-Scale over a Multi-Sites Grid using JavaSpace. Virginie Galtier. JavaSpace Overview. TupleSpace ( à la Linda) + Java OO + Jini services (transactions…) API : simple rich to build distributed

martinjuan
Download Presentation

Distributing a n-body Problem Algorithm at Large-Scale over a Multi-Sites Grid using JavaSpace

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Cracow Grid Workshop October, 18th 2006 Distributing a n-body Problem Algorithmat Large-Scale over a Multi-Sites Gridusing JavaSpace Virginie Galtier

  2. JavaSpace Overview TupleSpace (à la Linda) + Java OO + Jini services (transactions…) API : • simple • rich to build distributed applications with shared memory paradigm

  3. Objectives To investigate performance gains obtained when distributing a real and complete long-range data interaction application, on a large number of processors both on clusters and multi-sites grids. • impact of workers location ? • speed-up ? • evolution of best ExecTime/NbProc ratio when problem size increases ?

  4. N-Body Problem long-range data interaction

  5. 50 50 50 Worker Worker Worker 1 1 1 Master Distributed Algorithm homogeneous workers take write

  6. 50 50 50 Worker Worker Worker 1 1 1 1 1 1 Master Distributed Algorithm read find 1 group among 3 (instead of 2 bodies among 6)

  7. 50 50 50 Worker Worker Worker 1 1 1 1 1 1 1 1 Master Distributed Algorithm compute updated positions read computation/communication overlap

  8. 50 50 50 Worker Worker Worker 2 2 1 2 2 1 2 2 2 1 Master Distributed Algorithm write take free space from intermediate results

  9. 50 50 50 Worker Worker Worker 50 50 49 50 50 49 50 49 50 Master Distributed Algorithm

  10. Testbed

  11. Speed-up Study Speed-up Number of Workers

  12. Size-up Efficiency (%) Number of Workers rule-of-thumb to maintain a 90% efficiency: double P when N doubles

  13. Large Scale Extensibility Time / Body / Step (sec.) Number of Bodies O(N2)→O(N)

  14. Future Work • different kind of application? • different JavaSpace implementation? • compare with other Java-based middleware (ProActive) • influence of fault-tolerance mechanisms?

  15. Thank you

More Related