[Beowulf] longevity of 1U servers?

[Tim Cutts]

On 10 Feb 2009, at 12:10 am, Bill Broadley wrote:

>
> Seems like fans are the main problem with 1Us.   Pretty much  
> efficiency (air
> moved per watt) and noise (air moved per decibel) hugely favor  
> larger fans.
> I've seen 1U nodes that waste 30+ watts or so on moving air.
>
> Seems like a rear door with a few 18" diameter fans could easily  
> move more air
> than 40 1U systems with 240 1" fans in them.  Seems like large  
> savings in
> power and expense could be had by moving the fans to the rack  
> doors.  I've
> seen blade servers basically do this, alas at a large cost premium.

Exactly.  Consider the enormous blowers in an IBM BladeCenter chassis,  
there are only two fans for fourteen machines.

Blade servers do save money elsewhere;  *generally* they're much  
easier to deploy and administer than individual pizza boxes.

At least, they are when the lights out management firmware isn't an  
unspeakable crock.  Ahem.

Tim




-- 
 The Wellcome Trust Sanger Institute is operated by Genome Research 
 Limited, a charity registered in England with number 1021457 and a 
 company registered in England with number 2742969, whose registered 
 office is 215 Euston Road, London, NW1 2BE.