1 / 33

Parallel Computi ng Techniques

Parallel Computi ng Techniques. 1. Introduction 2. Parallel Machines 3. Clusters 4. Computational Grids 5. unGrid 6. Questions & Answers. SuperComputers  http://www.research.ibm.com/bluegene/. 1 . Introduction.  http://ungrid.unal.edu.co/ ScDeadProjects.html.

Download Presentation

Parallel Computi ng Techniques

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. Parallel Computing Techniques

  2. 1. Introduction2. Parallel Machines3. Clusters4. Computational Grids5. unGrid6. Questions & Answers

  3. SuperComputers http://www.research.ibm.com/bluegene/ 1. Introduction  http://ungrid.unal.edu.co/ScDeadProjects.html

  4. How to reduce processing time? 1. SuperComputers (101 - 104 µP,101 - 102 m$)2. Clusters (101 - 103 µP,50.000 - 101 m$)3. Grids (101 – 106..n µP,~0$) 1. Introduction

  5. Multitasking != Multiprocessing Multitasking: 1 CPU 1. Introduction CPU T1 T1 T1 T1 ... T2 T3 T2 T3 T2 T3 T2 T3 t ~10ms

  6. Multitasking != Multiprocessing Multiprocessing : 2+ CPUs 1. Introduction CPU1 T1 T3 T1 T3 ... T3 T1 T1 T3 T3 T1 T1 T3 CPU2 T2 T4 T2 T4 ... T4 T2 T2 T4 T4 T2 T2 T4 t

  7. Parallel Computing: 2+ CPUs 1. Introduction CPU1 T1 T1 T1 T1 ... T1 T1 T1 T1 T1 T1 T1 T1 CPU2 T1 T1 T1 T1 ... T1 T1 T1 T1 T1 T1 T1 T1 CPU3 T1 T1 T1 T1 ... T1 T1 T1 T1 T1 T1 T1 T1 ... t

  8. Distributed Computing: 2+ Machines 1. Introduction T2 T1 NET(LAN, MAN, WAN) T1 T2 T1

  9. 1. Introduction2. Parallel Machines3. Clusters4. Computational Grids5. unGrid6. Questions & Answers

  10. Flynn's Taxonomy A classification of computer architectures based on the number of streams of instructions and data:SISD / MISD / SIMD / MIMD 2. Parallel Machines

  11. SISDSingle Instruction Stream/Single Data Stream. sequential computer: PC (CPU+RAM) • MISDMultiple Instruction Stream/Single Data Stream. unusual 2. Parallel Machines

  12. SIMDSingle Instruction Stream/Multiple Data Stream. array processor or “vector processor” (Westinghouse, Fujitsy, NEC, Cray). all CPUs execute same instruction at same time. parallel done by different data • MIMDMultiple Instruction Stream/Multiple Data Stream. each cpu has own program and data. syncronization by communication 2. Parallel Machines

  13. Amdahl's Law (Gene Amdahl/IBM)“If it takes one man one minute to dig a post-hole then sixty men can dig it in one second” 2. Parallel Machines

  14. S : speedupF : sequential factionN : number of processors 2. Parallel Machines 1 S = F + (1-F)/N

  15. if N  infinite S = 1 / Fe.g. if F = 20%  Smax = 5 2. Parallel Machines

  16. 1. Introduction2. Parallel Machines3. Clusters4. Computational Grids5. unGrid6. Questions & Answers

  17. 3. Clusters LAN adm

  18. Distributed System: • Homogeneus platform (HW/SW) • Dedicated machines • Local area networks • Centralized administration 3. Clusters

  19. Gigabit ethernet cards/switches (60-90 MB/s) • RAM ~10GB/s (~102 faster) • Latency (50-100 µs)(~5 µs w/special cards) 3. Clusters

  20. 1. Introduction2. Parallel Machines3. Clusters4. Computational Grids5. unGrid6. Questions & Answers

  21. 4. Computational Grids LAN-MAN-WAN adm

  22. Distributed System: • Heterogeneus platform (HW/SW) • General purpose machines • LAN/MAN/WAN • Multiple admin domains(e.g. UN Bogotá campus/UN) 4. Computational Grids

  23. Specific Purpose Grids(seti@HOME, ...) • General Purpose Grids(unGrid, ...) 4. Computational Grids

  24. 10-100 ethernet cards/switches (1-10 MB/s) • Latency (10-200 ms) 4. Computational Grids

  25. 1. Introduction2. Parallel Machines3. Clusters4. Computational Grids5. unGrid6. Questions & Answers

  26. National University of Colombia: comprises several campuses across the country • The Bogotá campus is interconnected by a FDDI based net, where several LAN and VLAN coexist • Around 5000 computers interconnected by this net • Free CPU cycles are in the 60%-90% range 5. unGrid

  27. knowledge grid information grid computation/data grid • The unGrid Project attempts to construct a virtual supercomputer that takes profit from these CPU idle cycles 5. unGrid 

  28. Factors to consider when participating in a grid:. Privacy. Integrity. Transparency. Configuration/Uninstallation. Advantages 5. unGrid

  29. JavaSpaces 5. unGrid

  30. Master-Worker Pattern 5. unGrid

  31. 1. Introduction2. Parallel Machines3. Clusters4. Computational Grids5. unGrid6. Questions & Answers

  32. ? 6. Questions & Answers

  33. ungrid.unal.edu.co i n g r e s a ri n s t a l e u n G r i du t i l i c e u n G r i d

More Related