|
|
 | | From: | cigar | | Subject: | "Virtual" Cluster | | Date: | 21 Dec 2004 10:57:09 -0700 |
|
|
 | Has anyone done anything with clusters running on one host machines via
virtualpc 2004? I've created a 5-node "virtual cluster" (all are
RH9/96MB RAM) running within xp media center on a zd7000 laptop.
I'm curious because after running a very basic benchmark I got some
interesting results that I don't necessarily trust. But I also have no
clue what I'm doing, so I'm trying to figure out what I'm looking at.
Please see my site at http://liimpp.dwkuntz.net
My hardware is listed at http://liimpp.dwkuntz.net/beowulf/about.php
If anyone has any ideas/thoughts/suggestions, that would be awesome.
Thank you!
Dennis Kuntz
http://liimpp.dwkuntz.net
--
|
|
| | From: | GP | | Subject: | Re: "Virtual" Cluster | | Date: | 9 Jan 2005 22:08:56 -0800 |
|
|
| cigar wrote:
> Has anyone done anything with clusters running on one host machines via
> virtualpc 2004? I've created a 5-node "virtual cluster" (all are
> RH9/96MB RAM) running within xp media center on a zd7000 laptop.
>
> I'm curious because after running a very basic benchmark I got some
> interesting results that I don't necessarily trust. But I also have no
> clue what I'm doing, so I'm trying to figure out what I'm looking at.
>
> Please see my site at http://liimpp.dwkuntz.net
> My hardware is listed at http://liimpp.dwkuntz.net/beowulf/about.php
>
> If anyone has any ideas/thoughts/suggestions, that would be awesome.
> Thank you!
> Dennis Kuntz
> http://liimpp.dwkuntz.net
I agree, those results need explanation.
What, exactly, is making the MFLOPS measurements, and what is the time
base? Is each virtual machine running its own clock, which doesn't
change while another is running?
--
Greg Pfister
--
|
|
| | From: | cigar | | Subject: | Re: "Virtual" Cluster | | Date: | 13 Jan 2005 13:56:45 -0800 |
|
|
|
Each one is running it's own clock, using ntp. The ntp server being
used is a server at redhat.com
The MFLOP measurements are being done via the test pi calculation
program (that was modified by John Nordlie to be a crude measurement
tool). I used the first node as the control maching and simply did an
mpirun -np N flop where N is the number of processors on which I was
testing (1 through 5).
For the source to the MFLOP code, please see John Nordlie's site:
http://blizzard.rwic.und.edu/~nordlie/miniwulf/flop.c
--
------------ And now a word from our sponsor ----------------------
For a quality mail server, try SurgeMail, easy to install,
fast, efficient and reliable. Run a million users on a standard
PC running NT or Unix without running out of power, use the best!
---- See http://netwinsite.com/sponsor/sponsor_surgemail.htm ----
|
|
|
