Physical questions

Ken Chase math at velocet.ca
Mon Jan 13 23:19:58 PST 2003


On Tue, Jan 14, 2003 at 12:28:17AM -0400, Sam Daniel's all...
> Here's a slightly different approach which was successful on a small scale:
> 
> http://www.clustercompute.com/
> 
> Might not scale up to thousands of nodes, though.... :-)

or to 10 for that matter. When the middle one fails.

http://trooper.velocet.net/~math/images/133-3357_IMG.JPG

is what we did - custom cabinet, with brazeon mount points. Metal shelves.
4 nodes per shelf.

Without metal you get way too much RF. We found that stacking nodes
really close together vertically with no shielding caused many to not
boot properly (seemingly the most sensitive time) and many to crash
randomly (perhaps 10% of nodes every 2 weeks, mysterious crash).

Moving them to twice the distance between, my guess, reduced the crash
rate by a factor of 8ish (cube of distance? :)

With the metal shelves we had no such problems. The shelves in this cabinet
were also 9 inches apart, so its far enough even if they werent metal.

It was custom build by a kiosk builder, but it looked really great and
was comparable in cost to cases, but that many cases would never have fit
in the same volume of space...

/kc


> 
> Sam
> 
> Paul English wrote:
> > We are looking at building a second cluster and I have some physical 
> > installation questions. 
> > 
> > When I was at NASA goddard, I got a look at one of the Beowulf clusters 
> > there which had nodes which were just motherboards mounted on rails or a 
> > shelf which sat in the rack. Seems great to me to avoid purchasing a case 
> > if it isn't really needed - does anyone know any vendors for this sort of 
> > a mounting solution?  I found this one, but it looks like it includes the 
> > motherboards and the rack, while I'd like to just buy the shelves:
> >   http://www.kingstarusa.com/co-location/co-location_intelirack21.htm
> > I'd rather not fab my own, although I suppose that is an option. 
> > 
> > ... snip ...
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf

-- 
Ken Chase, math at velocet.ca  *  Velocet Communications Inc.  *  Toronto, CANADA 



More information about the Beowulf mailing list