The Art of Linux HPC Clusters

The Art of Linux HPC Clusters:
Building Machines Around Problems

A proposed book by Douglas Eadline, Ph.D. discusses the art of designing a successful Linux HPC cluster. The book looks at many of the over looked, but important, issues involved with building an HPC cluster. Uses can expect to gain a good fundamental understanding of essential cluster technologies and methods.

Planned publication date is sometime in 2010.

Check back for more information and publication date. Questions or Inquiries

Notes From the Author:

I have probably read all of the available Linux cluster books(that includes the MPI and PVM books as well). In reading these books I have found a large amount of good information about clustering hardware and software. The one thing that seems to be missing is a discussion of the "craft" or "art" of Linux HPC clustering.

Indeed, over the years, I have found that building effective Linux HPC clusters is an art form. As such, this book will not tell you what interconnect to buy or what processor to use. (Coming soon as well!) It will tell you, in a concise fashion, the issues that you need to think about before you spend your time and money building a machine around your problem(s). The benefit to you is less wasted time and money and a faster "time to solution" for your problem(s). My current chapter outline is as follows. Check back for availability and publication information.

  1. Introduction To An Art Form
    • Goals
    • What You Can Expect From This Book
  2. The HPC Storm
    • A Perfect Storm
    • After The Storm
    • Problems First, Machines Second
  3. Great Expectations
    • Low-cost high-performance
    • Urban Legends
    • A Taste of Reality
    • What Really Matters
  4. A Paradigm Shift
    • The commoditization of computing
    • Building machines around problems.
    • Best tools are a pencil and paper
  5. The Fundamentals
    • It's the Software, Stupid
    • What Was Hard just go Harder
    • Understanding Concurrency
    • Bye the way, Multi-core is Parallel Computing
  6. The Open Approach
    • There is no Free Beer (Darn)
    • Linux On Clusters
  7. Fans and Wires: The Hardware
    • We Take What We Are Given
    • Sharing Memory with SMP
    • Copying Memory with Interconnects
    • Infrastructure Options
  8. The Plumbing: The Software
    • Image based systems
    • Global Processes
    • Sending Messages
  9. Coping with Complexity: Administration
    • Monitoring the Monitoring
    • The Tools You Will Need
    • The Tools You May Not Need
    • The Tools You Will Want
  10. A Word About File Systems
    • Welcome to the real World
    • Some Options
  11. Your Next Steps
    • Turn-Key Applications
    • Writing Software
    • World Domination
  12. Resources

Check back for more information and publication date. I'm also interested in your comments or suggestons.

    Search

    Feedburner

    Login Form

    Who's Online

    We have 154 guests and no members online

    Share The Bananas

    Worldwide Front Page Visits

    Monkey Stats

    Google PageRank module - Camelpark SEO centrum

    Reader Poll

    Do you have and interest in Big Data and HPC

    No, I have no need for working with large datasets - 5.4%
    Yes, I have more data than I can handle - 35.1%
    I'm not sure what Big Data can bring to my HPC efforts, I would like to learn more - 37.8%
    I already use Big Data Tools on my cluster! - 21.6%

    Total votes: 37
    The voting for this poll has ended on: 30 Jun 2014 - 00:00

    Creative Commons License
    ©2005-2012 Copyright Seagrove LLC, Some rights reserved. Except where otherwise noted, this site is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 2.5 License. The Cluster Monkey Logo and Monkey Character are Trademarks of Seagrove LLC.