[Beowulf] Re: ECC Memory and Job Failures (Huw Lynes)

Robert G. Brown rgb at phy.duke.edu
Fri Apr 24 05:40:10 PDT 2009


On Fri, 24 Apr 2009, John Hearns wrote:

> 2009/4/24 Greg Lindahl <lindahl at pbm.com>:
>> On Thu, Apr 23, 2009 at 10:20:02PM -0400, Robert G. Brown wrote:
>>
>>
>> On clusters of ~300 nodes over burnin times of weeks, I was easily
>> able to see the difference between sea level, Boulder, and
>> Albuquerque, with 2000's-era memory.
>
> Hmmmm. All we need is IEEE-1588 precision time links between clusters
> and we've got a heck of a wide area cosmic ray telescope.

Aw, you're stealing one of my ideas, except it involves cell phone
towers.  Add a single inexpensive compute box with an attached radio and
some e.g. a2d to every cell phone tower in the country (or better, the
world) and the entire populated planet would become one truly enormous
radiotelescope.  Telescope Earth, the world's "largest" beowulf project.
A project made simpler by virtue of the fact that most towers now are
virtual GPS stations as is and hence are precisely located in space and
time

Just doing this across North America would be astounding.  A baseline of
some 5000 km, full field northern hemisphere view.  Using waves in the
meter to millimeter range, an angular resolution of 10^-6 to 10^-9
radians.  That's very close to being able to resolve nearby suns as
disks and not points in the millimeter range.

Truthfully, nobody should ever build any other style of radiotelescope
again.

I've tried to get an undergrad to actually tackle this one as a special
project or senior thesis, with no luck.  One looked at it, but it needs
somebody with time and energy to build a handful of boxes and actually
talk tower owners into pro bono access to the tower antenna signal and a
(free) chunk of fiber BW to retrieve results, probably during non-peak
usage during the night.

Now for the good part -- my plan was to make that single inexpensive box
basically a small PC, one with a TB or so of hard storage and a fiber
connection back to a collector cluster and leeching GPS-grade time data
from the tower.  Using e.g. ECC error rates as a secondary radiation
detector (or, sigh, adding e.g. a USB interfaced actual detector
although that isn't in the spirit of the thing:-) and one can do lots of
interesting cosmological things -- look for particle bursts lagging
radio bursts in a highly directional cone, basically picking up things
like supernovae signatures.

rgb

> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
>

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu





More information about the Beowulf mailing list