[lug] server spec

[Ed Moxley]

On Tue, 2007-06-12 at 23:39, Sean Reifschneider wrote:
> On Mon, Jun 11, 2007 at 12:26:09PM -0600, Nate Duehr wrote:
> >The heat problems of 1RU are many, and I've seen other (better)  
> >machines like Dell, HP, and IBM machines all die untimely deaths  
> 
> I'd guess it's a problem with your environment.  With modern systems, you
> have to be extremely careful about the aerodynamics of the server room.
> You can't just stick one thermostat in there and if it reads 70 degrees
> assume everything is fine.  You also have to be very careful about knowing
> where exhaust is going and where intake is coming from.
> 
> Putting an amp or two through 1.75 inches of rack space means you have to
> move a lot of air.  Is the intake air 70 degrees, or is it getting recycled
> from the exhaust of other machines?  We have cabinets and cabinets of
> 1U machines in our managed hosting facility, largely Supermicro, some
> 4 to 5 years old.
> 
> I would *NOT* agree that 1U machines "all die untimely deaths".  We have
> few if any problems with them.  We don't do anything unusual to the systems
> with respect to cooling, we just engineer it so that they are getting good
> intake and pretty much can forget about them...
> 
> >machines use some kind of (really loud) squirrel cage fan setup --  
> >the SuperMicro's I've used had multiple little 1" fans, and they die  
> >off pretty quickly.  A couple of years, tops.   I guess the idea is  
> 
> Supermicro varies the type of fan from chassis to chassis, depending on the
> needs.  We have a number of them with squirrel cage fans, many others that
> have the small traditional fans, usually the stacked fans with one running
> X RPM connected to another doing 2X RPM, to move a lot of air.  I haven't
> had any problems with the fans of either type on our Supermicros.
> 
> Fans can vary in quality quite a lot.  I've always tried to get high
> quality fans, but even then I find some that last less than a year before
> they seize up, others I've had running for 5+ years and were still going
> strong.
> 
> Sean

My experience has been similar to Sean's.  We have around 300 1U nodes
with dual 3.2 GHZ EM64T CPUs in the same room with 700+ 2U nodes with
dual 2.2 GHZ Xeons.  All are Supermicro and the only heat problems we've
had are from when we lose cooling to the room.
-- 
Ed Moxley <ed at moxleynet.com>