Adaptive parallel computing over distributed military computing infrastructures <br />RiturajKumar<br />Center of Artifici...
2<br />Contents<br /><ul><li>Introduction
Adaptive Parallel Computing
Approaches for APC
MPI + DDS
Hadoop
Conclusions</li></li></ul><li>3<br />Introduction<br /><ul><li>Net-Centric Paradigm of warfare needs highly compute intens...
Parallel Computing is the only feasible solution that would guarantee reduced response time.
Tactical deployment would not amenable for –
Large Clusters in the field
Back-hauling
Intelligent use of the existing spare compute capacity of the computing devices within the cloud is essential.</li></li></...
Upcoming SlideShare
Loading in …5
×

Apache Hadoop India Summit 2011 talk "Adaptive Parallel Computing over Distributed Military Computing Infrastructures" by Rituraj Kumar

1,632 views
1,574 views

Published on

1 Comment
0 Likes
Statistics
Notes
  • what was this presentation all about? announcing the title itself must take like 29min. at least i couldn't make a head or tail of it.
       Reply 
    Are you sure you want to  Yes  No
    Your message goes here
  • Be the first to like this

No Downloads
Views
Total views
1,632
On SlideShare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
42
Comments
1
Likes
0
Embeds 0
No embeds

No notes for slide

Apache Hadoop India Summit 2011 talk "Adaptive Parallel Computing over Distributed Military Computing Infrastructures" by Rituraj Kumar

  1. 1. Adaptive parallel computing over distributed military computing infrastructures <br />RiturajKumar<br />Center of Artificial Intelligence and Robotics – DRDO Lab<br />
  2. 2. 2<br />Contents<br /><ul><li>Introduction
  3. 3. Adaptive Parallel Computing
  4. 4. Approaches for APC
  5. 5. MPI + DDS
  6. 6. Hadoop
  7. 7. Conclusions</li></li></ul><li>3<br />Introduction<br /><ul><li>Net-Centric Paradigm of warfare needs highly compute intensive operations to be performed.
  8. 8. Parallel Computing is the only feasible solution that would guarantee reduced response time.
  9. 9. Tactical deployment would not amenable for –
  10. 10. Large Clusters in the field
  11. 11. Back-hauling
  12. 12. Intelligent use of the existing spare compute capacity of the computing devices within the cloud is essential.</li></li></ul><li>4<br />Adaptive Parallel Computing<br />Cloud of Heterogeneous Computing Devices<br />MILCOM<br />
  13. 13. 5<br />Adaptive Parallel Computing<br />Amdahls Law<br />The Amdahl's law is concerned with the speedup achievable from an improvement to a computation that affects a proportion P of that computation where the improvement has a speedup of S. <br />Where,<br />S = Speedup<br />P = Parallel fraction of Program<br />N = Number of Processors<br />If 95% of the program can be parallelized, the theoretical maximum speedup using parallel computing would be 20×, no matter how many processors are used. <br />
  14. 14. 6<br />Adaptive Parallel Computing<br />Cost for parallel Computation = Computation Cost + Serialization Cost<br />
  15. 15. 7<br />APC Approaches<br />MPI + DDS<br />MPI Code:<br /><ul><li>Parallel implementation of </li></ul> complex mathematical models<br />MPI Controller:<br /><ul><li>Execution of parallel MPI code</li></ul> over distributed network. <br />DDS:<br /><ul><li>Reliable Communication over</li></ul> challenged network.<br />DMI:<br /><ul><li>Identification of computing </li></ul> nodes in the distributed <br /> network.<br />MPI Code<br />MPI <br />Controller<br />Dynamic Membership Identifier<br />DDS<br />
  16. 16. 8<br />APC Approaches<br />MPI + DDS<br />Advantage:<br /><ul><li>Well-known MPI API Framework.
  17. 17. DDS provides assurance of reliable communication between nodes.</li></ul>Disadvantage:<br /><ul><li>Needs a wrapper for converting MPI calls to DDS framework.
  18. 18. Performance degradation because of the wrapper.</li></li></ul><li>9<br />APC Approaches<br />HADOOP<br /><ul><li>Map/Reduce can be effectively used for parallelization
  19. 19. Used extensively in varieties of information systems.
  20. 20. Java Based
  21. 21. Designed to handle large data sets</li></li></ul><li>10<br />APC Approaches<br />Hadoop<br />Advantage:<br /><ul><li>Highly Scalable
  22. 22. Excellent fault tolerance capabilities
  23. 23. Good hardware interoperability</li></ul>Current challenges:<br /><ul><li>Requirement of slight architectural changes
  24. 24. Currently not suitable for resource constraint hardware</li></li></ul><li>11<br />Conclusion<br /><ul><li>Military domain requires reliable parallel computing infrastructure over disadvantaged communication network.
  25. 25. Dynamic topology poses great challenge for computing infrastructure.
  26. 26. MPI framework has few disadvantages
  27. 27. Hadoop is a promising candidate in these conditions.</li></li></ul><li>12<br />Thank You<br />

×