On Monday, November 10, 2003, at 10:07 AM, Kelvin Chu wrote:
> While I bemoan the fact that we don't have a BigMac computing cluster,
> I'd settle for this, too.
> November 10, 2003
> Europe Exceeds U.S. in Refining Grid Computing
> By JOHN MARKOFF and JENNIFER L. SCHENKER
> When the Swiss-based pharmaceutical giant Novartis needed a new
> supercomputer for designing drugs, the company found it already had
> one. It was hidden in the unused computing power the company had
> available in the thousands of PC's that were already being used in its
Well, funny you should say that...
I just got back from the NSF Second NMI Integration Testbed Workshop on
"Experiences in Middleware Deployment"
where one of the topics was grids.
What you (and we) want is the Globus Toolkit
One of the interesting "Experiences in Middleware Deployment" was the
USC Case Study presented by Shelley Henderson. They had a public
computing lab full of Sun workstations which they incorporated into
their grid. Students, unaware (or uncaring) of this fact, would come
into the lab and try to login. The login process was slowed -- just a
bit but enough -- by the grid software (and the distributed processes
executing on the grid), so the impatient students would choose to
reboot the machines instead of wait. This behavior was not seen by the
USC researchers as especially helpful to their calculations.
USC has some heavy iron in their grid, too: see
http://www.usc.edu/hpcc/hpc.html , http://www.usc.edu/hpcc/almaak.html
So, what do we learn from the USC experience?
o It isn't the technology. It's the turf and politics of rounding up
enough nodes to make the grid worth the effort
o having a couple of monster machines in the mix doesn't hurt
If you are serious about wanting a grid, let's talk. I've got GT3
building on my Mac OSX test box -- a 266 MHz G3 . If I can build it
there (whoops -- maybe I can't -- package
org.globus.ogsa.impl.security.authentication.service does not exist),
it can be built anywhere. Find me another couple dozen machines and
let's see what we can assemble.
| Wesley Alan Wright <mailto:[log in to unmask]> |
| Academic Computing Services __0__ |
| Room 238 Waterman Building / \ | \ |
| University of Vermont \77 |
| Burlington, Vermont 05405-0160 USA. \\ http://www.uvm.edu/skivt-l |
| Voice:802-656-1254 FAX:802-656-0872 vv |
| aim:goim?screenname=maddogskideath http://weasel.uvm.edu/ |