[rescue] Deciding how to place machines in a rack...

[Robert Darlington]

It absolutely holds true for 1U servers.  Dead air is insulation.  Or
the worst case scenario is when you have a slightly higher pressure in
the back (hot air) leaking around the servers back to the cold side.
This is great when you're trying to warm up a car, but not so good for
your servers.

-Bob

On Wed, Mar 5, 2008 at 3:26 PM, Brian Deloria <bdeloria at gmail.com> wrote:
> I can see this making sense and being true for Steve's post definately with
>  some proliants coming in at 7u.
>
>  Does this also hold true for 1u servers, something that's higher density
>  etc.?
>
>  That was always where I'd see it the most.
>
>  If we're talking about just relying on the fans that are internal to the
>  machines does having spaces matter?  i.e. no forced air up through the floor
>  etc.
>
>  I know most manufacturers when you replace a drive on a machine want you
>  either to replace it immediately with another or with a filler.  The ibm
>  bladecenter equipment I worked with years ago even went so far as there were
>  fingers that would spring up from the bottom and down from the top to impede
>  air flow through the voids in the chassis when a blade was removed.
>
>  I lost a good bit of hearing from those fan's too.  any computer that has
>  "Accoustic Attenuator" as an option is pretty freakin loud.
>
>
>  Thanks,
>
>  Brian
>
>  On Wed, Mar 5, 2008 at 3:52 PM, Robert Darlington <rdarlington at gmail.com>
>  wrote:
>
>
>
>  > Oh man, everyone at a previous job (national lab) would insist on
>  > putting spaces between servers until I ran a simulation with COSMOS
>  > FlowWorks and showed them their mistakes.  This is terrible for
>  > cooling systems.  You guys are absolutely right about dead air space
>  > insulating the heck out of machines.  The only way this makes sense is
>  > if you have forced air coming from the front of the rack, such as a
>  > fan or chiller unit.
>  >
>  > Nevermind you waste a lot of rack space when you leave gaps between
>  > servers.
>  >
>  > -Bob
>  >
>  > On Wed, Mar 5, 2008 at 2:40 PM, Steve Hatle <shatle at nfldinet.com> wrote:
>  > > On 3/5/08 3:21 PM, "Patrick Finnegan" <pat at computer-refuge.org> wrote:
>  > >
>  > >  > On Wednesday 05 March 2008, Earl Baugh wrote:
>  > >  >> I'll space each machine 1U apart, but I have a question regarding
>  > >  >> that.
>  > >  >
>  > >  > FWIW, I'd suggest making the machines as contiguous as possible,
>  > >
>  > >
>  > >  I think I'd agree with Pat on this one; I can't see any advantage of
>  > getting
>  > >  air between the systems. The little cooling you would get would be
>  > radiant,
>  > >  and unless you are forcing air through those spaces, it would be
>  > minimal.
>  > >
>  > >  The machines (should) be designed to pull the air they need through
>  > them,
>  > >  whether from the front or side as the case may be.
>  > >
>  > >  As far as using the plexi door, we did an experiment with some of our
>  > gear
>  > >  at a previous job. These machines (Proliants) drew air from the front,
>  > and
>  > >  exhausted from the rear. Air was forced up from the floor under and in
>  > front
>  > >  of the racks. No top fans in the racks.
>  > >
>  > >  We used the plexi doors on the front of the racks and used the built-in
>  > temp
>  > >  monitoring in the machines to get a baseline of internal temps. The
>  > plexi
>  > >  doors did have some venting in them, but not much.
>  > >
>  > >  We then switched them for the mesh doors that went on the back of the
>  > racks,
>  > >  and left the backs open. The internal temps of the machines went down
>  > >  between 0 and 10 degrees F, without any other changes to the
>  > environment,
>  > >  etc.
>  > >
>  > >  These were 42U racks with the machines and disk trays stacked from
>  > bottom to
>  > >  top with no spaces in between. Never even considered that airspace
>  > between
>  > >  would be necessary- besides, I needed every U of rack space.
>  > >
>  > >  So- I don't know how that might effect your decision. I'd say if you
>  > like
>  > >  the plexi door, then use it. If you think 10 degrees is worth it, then
>  > you
>  > >  can modify the door to suit.
>  > >
>  > >  Besides, the blinky lights are brighter without the door if it's
>  > >  colored/smoked plexi :-)
>  > >
>  > >  Steve
>  > >
>  > >
>  > > _______________________________________________
>  > >  rescue list - http://www.sunhelp.org/mailman/listinfo/rescue
>  > _______________________________________________
>  > rescue list - http://www.sunhelp.org/mailman/listinfo/rescue
>  _______________________________________________
>  rescue list - http://www.sunhelp.org/mailman/listinfo/rescue