[Beowulf] Re: Disks fail at what acceleration? (Jim Lux)
Robert G. Brown
rgb at phy.duke.edu
Thu Mar 23 08:54:49 PST 2006
On Thu, 23 Mar 2006, David Mathog wrote:
> Jim Lux wrote
>> At 01:24 PM 3/22/2006, David Mathog wrote:
>>> Anybody know how many g's a typical disk can withstand before it fails?
>> A lot more than 1g...
>> But, since you want "real numbers"
>> looks like 63 G shocks over 2msec, operating... 350 G over 1 msec, non-op.
> Ok. So it looks like a bolted rack should be ok so long as it moves
> smoothly with the ground motions. On the other hand a bunch of
> computers loosely restrained on shelves might or might not
> exceed these limits when they bounce off the shelving sides or the back
> wall. Ditto for a desktop computer. These collisions are going to
> be more elastic than inelastic but there is some give in all of the
> materials so the peak Gs even in that case probably won't exceed that
> limit. Unless of course one falls off the shelf and onto the
> concrete floor...
You should look into raised floor designs and special racks to withstand
earthquakes, depending on how serious you are (read, how much want to
spend). I did a bunch of online research for this last year sometime in
response to a list thread and some local needs and there are actually
shock-absorber extra-strong risers and pedastals out there for
earthquake-proofing server/cluster environments. For a cost, naturally,
but it tends to be a one-time cost. There is actually a bunch of data
out there on retrospective analyses of e.g. some relatively famous
failures of raised floor and other infrastructure in some of the
semi-major earthquakes of the last decade and a half, where some designs
that were supposed to work did not and others that weren't really
designed for it did OK.
Remember, it isn't just "g forces" -- earthquakes cause the ground to
vibrate. If a vibration happens to have a strong frequency component in
a band that is resonant with some part of a disk, it can cause failure
even though the peak acceleration of the ground is relatively low. E.g.
- disk heads cannot be too strongly damped inside because damping equals
resistance equals both heat and a cost in speed. Shock mounting of the
racks can reduce the amount of vibration that is transmitted directly
from the ground up through rigid components. Or I recall over the years
seeing and hearing of disks that had rubber shock-absorbing bushings on
their internal mounts, to minimize vibration transfer within the actual
mounting frame. This also helps to minimize operational noise the other
way, keeping vibrations from the disk from coupling to the acoustical
cavity of the box itself.
However, the BULK of the damage I recall being done by earthquakes
occurred with sections of a raised floor just collapsed, dropping whole
racks of equipment some feet straight down to the ground, shorting out
wires, crushing networks, and even tipping over to smash a whole strip
of flooring and its associated stuff out. It took them weeks to just
untangle the the physical wreckage at one server room that was
"supposed" to be out only a day at most. GIYF here.
> David Mathog
> mathog at caltech.edu
> Manager, Sequence Analysis Facility, Biology Division, Caltech
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
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
More information about the Beowulf