 |
|
DAILY NEWS AND INFORMATION
FOR THE GLOBAL GRID COMMUNITY /
|
Special Features:
CONCLUSION: 7 STEPS TO SUCCESSFUL
GRID COMPUTING
By Brooklin J. Gore, Senior Fellow, Micron Technology
Inc
This report along with two other new, groundbreaking Grid reports from IDC
and
the Economic Strategy Institute will be released at Gt'04 and included with
conference registration. www.gridtoday.com/04/conference/index.html
This is the final installment of a three-part series derived from the
report
"Enterprise Grids For General Purpose Computing" by Brooklin J. Gore.
Web services, utility computing, .NET, CPU harvesting and distributed
computing are just a few of the technologies that fall under the Grid
computing umbrella. Gt04 -- a premiere enterprise Grid computing conference
targeting industrial and commercial users -- will gather experts, and outline
strategies and road maps for Grid deployment. For more information, visit
www.gt04.com.
Grid computing is here!
Lack of developer trust and reluctance to change work to slow new
technology
deployment. Moreover, rightly so. If your Grid is not as robust and reliable
as your traditional compute platforms, your Grid efforts will become
untrusted. It is hard enough to get people to take a chance on something new,
but when they do, you must ensure them a positive experience. The low hanging
fruit, education and integration keys (5, 6 and 7) above come to play here.
Start with developers and applications that can be flexible regarding uptime
and use your experience with these to build a solid production class Grid with
a well-integrated support model.
Understand thick client and master worker program models of software
development. While many "off the shelf" programs will run on the Grid, in the
grand scheme of things the Grid is a jungle. It can be a tangle of
misconfigured machines, machines that lie about their capabilities and
machines that just go away while running a job. Grid software does a great job
of dealing with these challenges, but ultimately for maximum reliability and
use of resources Grid programs should be able to handle a variety of error
conditions they might not be used to. Further, Grid jobs must learn to be
nice. Nobody worries about a single ant in his or her home -- it is the colony
we worry about. Well, Grid jobs generally appear in colonies and as such need
to behave. They run on borrowed hardware and shouldn't make a mess of it.
Collectively they can overwhelm a database server and should learn when to
wait to connect. These are just a few challenges of Grid programming. It is
more difficult to get Grid developers to adopt this new way of thinking than
for them to actually implement these new ideas. We reiterate keys 4 and 6 in
articulating the value proposition and education.
Not many organizations have experience managing collections of ten thousand
machines with hundreds of users and millions of jobs per day in a single
environment. Will Grid software be able to ensure the right user's jobs are
getting the resources they need at these orders of magnitude? By the end of
2004, we expect our largest collection to be right around 10,000 machines.
This author has as much pioneering spirit as the next, but he is not excited
to be the first guy to discover a new flesh-eating virus at the heart of the
Amazon. Are you? Therefore, here's a challenge to the Grid Software folks:
we're using this stuff in ways you may have never imagined and we love what we
see -- please listen to our needs and be prepared to meet them.
These challenges can and will be overcome to yield great rewards. In August
of
2003, over 96,000 hours of work were completed on the Grid. Assuming $1,000 PC
equivalent systems performing work full time (720 hours a month), this works
out to 133 machines -- that didn't have to be purchase. That's a $133,000
savings. Of course, additional value is provided by making decisions on better
data that is delivered more quickly. Costs for training, pool administration
and developer support have been negligible.
Next Steps
If you do not have an active Grid Computing program in your organization,
it
is time to start one. If you have a fledgling program, hopefully, some of the
tips provided in this article will accelerate your success and that the
application examples have provided new incentives for the broad applicability
of your Grid. If your Grid program is a great success you've probably been
smiling a lot while reading this article as I'm sure much of it is familiar
ground to you by now.
We continue to grow our machine collections aggressively and at the same
time
work with an increasing number of developers who want to Grid-enable their
software. We also continue spreading the value proposition of grid computing
and educating administrators and developers worldwide. Most importantly, we
are working with our Grid software supplier to ensure they are aware of our
current usage and future plans so they may anticipate potential problems
before they manifest themselves in a production environment.
Acknowledgements
The Author would like to thank the following people for their pioneering
spirit and efforts to make our Grid Computing efforts at Micron a success:
Mike Dorough, Sam Evans, Todd Evans, Eric Fletcher, Chin Le, Michael Lee, Tim
Long, Joe Lundgren, Ed Mahoney, Bryan Mann, Melissa Norgard, Adriana Sanchez,
Mike Steele, Mike Urizar, Doug Warner and Bridger Wood.
Grid Primer
A wide selection of Grid software is available from open source and
commercial
providers. Condor from the University of Wisconsin, Sun ONE Grid Engine from
Sun Microsystems and LSF from Platform Computing are examples that span the
open source to commercial range.
A well-deployed and functioning Grid provides the following benefits:
- Lowers computing costs by more efficiently using existing
resources.
- Increases application reliability by dynamically directing jobs to
available resources.
- Improves decision making by delivering huge amounts of raw computing
power.
About Brooklin J. Gore
Brooklin Gore has been researching and implementing enterprise grid
technologies for the past three years to create Micron's global grid
infrastructure which runs over 15 production applications today. Brooklin has
been with Micron for 16 years. In that time he served as a product engineer,
Computer Aided Design group manager, network manager and General Manager of
Micron's Internet Services Division. Brooklin has been issued several US
patents and is a Senior Member of the IEEE. He holds Bachelor of Science
degrees in Computer Science and Electrical Engineering from the University of
Idaho and a Masters of Science in Computer Science from the National
Technological University.
|