slide01 intro

Upload: -

Post on 07-Apr-2018

221 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/6/2019 Slide01 Intro

    1/31

    517 454, 517 441:

    Parallel and DistributedComputing

    Apisake [email protected]

    mailto:[email protected]:[email protected]:[email protected]
  • 8/6/2019 Slide01 Intro

    2/31

    What is Parallel Computing?

  • 8/6/2019 Slide01 Intro

    3/31

  • 8/6/2019 Slide01 Intro

    4/31

    Parallel Computing

    In the simplest sense, it is the simultaneoususe of multiple computing resources tosolve a computational problem.

    Problem

  • 8/6/2019 Slide01 Intro

    5/31

    The computer resources

    +

  • 8/6/2019 Slide01 Intro

    6/31

    The computational problem

    Demonstrates characteristics such as theability to be:

    Broken apart into discrete pieces of workthat can be solved simultaneously

    Execute multiple program instructions atany moment of time

    Solved in less time with multiple computeresources than with a single computerresource

  • 8/6/2019 Slide01 Intro

    7/31

    Think about

    Natural World!! Parallel computing attempts to emulate

    what has always been the state of affairs inthe natural world

    many complex, interrelated eventshappening at the same time, yet within asequence

  • 8/6/2019 Slide01 Intro

    8/31

    Grand Challenge Problems

    Weather and Climate

  • 8/6/2019 Slide01 Intro

    9/31

    Grand Challenge Problems

    Biological, Human Genome

  • 8/6/2019 Slide01 Intro

    10/31

    Grand Challenge Problems

    Geological, Seismic Activity

  • 8/6/2019 Slide01 Intro

    11/31

  • 8/6/2019 Slide01 Intro

    12/31

    Grand Challenge Problems

    Search for Extraterrestrial Intelligence

  • 8/6/2019 Slide01 Intro

    13/31

    Grand Challenge Problems

    Search for Extraterrestrial Intelligence

  • 8/6/2019 Slide01 Intro

    14/31

    Commercial

    ApplicationsThe Driving Force

    ..need faster computers for processing oflarge amount of data in sophisticated ways

  • 8/6/2019 Slide01 Intro

    15/31

    parallel databases, data mining oil exploration web search engine, web-based business

    services

    computer-aided diagnosis in medicine management of national & multi-national

    corporations

    advanced graphics and virtual reality and a lot more...

  • 8/6/2019 Slide01 Intro

    16/31

    Why UseParallel Computing?

  • 8/6/2019 Slide01 Intro

    17/31

    Primary Reasons

    Save time - wall clock time Solve large problems

    Provide concurrency

  • 8/6/2019 Slide01 Intro

    18/31

    Other Reasons...

    Taking advantage of non-local resources Cost-savings

    Overcoming memory constraints

  • 8/6/2019 Slide01 Intro

    19/31

    Limits to Serial Computing

    Transmission speeds Limit to miniaturization

    Economic limitations

  • 8/6/2019 Slide01 Intro

    20/31

  • 8/6/2019 Slide01 Intro

    21/31

  • 8/6/2019 Slide01 Intro

    22/31

    Whos Doing Parallel Computing?

    17%

    1%

    3%

    57%

    20%

    2%

    Academic Classified GovernmentIndustry Research Vendor

    Segments share for 11/2007

  • 8/6/2019 Slide01 Intro

    23/31

    What are they using it for?

    0

    75

    150

    225

    Aerospace

    Benchmarking

    CFD

    Database

    Economics

    Environment

    Geophysics

    InformationService

    Manufacturing

    Media

    Research

    Software

    Transportation

    WWW

    DigitalMedia

    Gaming

    Application Area share for 11/2007

  • 8/6/2019 Slide01 Intro

    24/31

    Concepts

    &Terminology

  • 8/6/2019 Slide01 Intro

    25/31

    von Neumann Architecture

    Store-program concept - the CPU executesa stored program that specifies a sequenceof read & write operations on the memory

    CPU

    Memory

    fetch execute(manipulate data asprogrammed)

  • 8/6/2019 Slide01 Intro

    26/31

    Basic Design

    Memory is used to store both program anddata instruction

    Program instructions are coded data whichtell the computer to do something

    Data is simply information to be used bythe program

    A CPU gets instructions and/or data frommemory, decodes the instructions, and thensequentially perform them

  • 8/6/2019 Slide01 Intro

    27/31

    Flynns Classical Taxonomy

    SingleInstruction

    MultipleInstruction

    SingleData

    MultipleData

  • 8/6/2019 Slide01 Intro

    28/31

    Single Instruction,

    Single Data (SISD) A serial computer

    Single instruction: oneinstruction stream /clock cycle

    Single data: one datastream / clock cycle Deterministic

    execution

    time

  • 8/6/2019 Slide01 Intro

    29/31

    Single Instruction,

    Multi le Data SIMD

    time

    P1 P2 Pn

  • 8/6/2019 Slide01 Intro

    30/31

    Multiple Instruction,

    Single Data (MISD)

    time

    P1 P2 Pn

  • 8/6/2019 Slide01 Intro

    31/31

    Multiple Instruction,

    Multi le Data MIMD

    time

    P1 P2 Pn