[Beowulf] Earthquakes and raised floors...

Robert G. Brown rgb at phy.duke.edu
Mon Jan 9 03:01:16 EST 2006


On Sun, 8 Jan 2006, Joel Jaeggli wrote:

>> However (and this isn't argument or rhetoric either, just a plea for
>> enlightenment) -- just what is APC trying to sell with the white paper?
>> I know they sell a variety of rackware and power equipment, but what
>> have they got against raised floors (specifically, what sales advantage
>> do they gain if a client doesn't use a raised floor)?
>
> They make dedicated airhandling units for really hot racks.
>
> They also offer a rather upmarket datacenter design and deployment service 
> presumably they have some customers that want or think they want raised 
> floors so a white paper that says you don't really need this thing that we 
> don't provide seems likely.
>
> Not that I have anything against apc, I've got all sorts of their products 
> deployed. but I do take their marketing materials with a grain of salt.

Fair enough.  Still, though, a lot of the points they make seem valid.
A raised floor wasn't even possible to contemplate in our server/cluster
room -- too low a ceiling, too much crap up there already that couldn't
be moved without spending even more money than we spent (and we spent a
LOT renovating it, as AC's are expensive).  So their observation that
limited overhead space argues against a raised floor sure worked for us
even without our reading it from a white paper or anything;-)

In case my original presentation wasn't clear, though, I wasn't
advancing MY opinion when I said the "raised floors are obsolete" thing
-- I was just summarizing the white paper, for the specific purpose of
pointing out that according to MULTIPLE information sites on the web
(including this white paper), raised floors and earthquakes are not
happy partners.  They have a higher probability of failure than the
concrete floor the raised floor sits on, and if they fail the failure is
likely catastrophic for the equipment involved.  "Earthquake proof" also
appears to be more a matter of marketing hype than reality, as well --
postmortems of datacenters in Kobe and the Bay area after major
earthquakes showed that plenty of supposedly earthquate proof floors
collapsed anyway, and that those collapses led to LONG downtime and lots
of expensive equipment failure as it isn't easy to basically take apart
the floor and cabling, lift out the smashed equipment, rebuild the
floor, and replace the equipment, safely and quickly.

Cooling failure failure or water leakage failure may well be larger
risks, I don't know, and THESE can probably be mitigated in various ways
(as can the degree of earthquake-proofness be varied) making the issue
still more complex.

The main conclusion one might draw from the whole discussion is that the
issue of raised floor or no raised floor isn't one with any single
"right" answer.  YMMV.  There are different costs, benefits, risks.  On
the east coast we don't get a lot of earthquakes (although when we get
one it can be a doozy).  On the west coast there are lots of small
earthquakes and relatively frequent big ones.  Some spaces have limited
ceiling overhead.  Others have no way to put in an access ramp.  Still
others might have warehouse sized spaces where those things are no
problem at all.  You should feel free to put in a raised floor if it
meets your needs and specifications and its advantages appeal, but you
should feel free to NOT put in a raised floor even though they were
traditionally used in lots of data centers in the past if they DON'T
make sense for your particular needs and environment.

YMMV.  We knew that....:-)

    rgb

-- 
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


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



More information about the Beowulf mailing list