Not many people are aware, but I was surprised, after discussing with some engineers a few years ago,
that the physical vibrations of equipment in their racks can actually cause power increases
and performance issues purely as a result of the vibrations interfering with disk seek times
and so on. The sympathetic vibrations of the disk and media is actually perceptible
when tests are run. As the vibrations are constant, this will affect the way the heads on the
disk access the data. When the vibrations were suppressed, disk seeks improved commensurately and a noted decrease in power consumption and response time was noted.
Multiply these factors when taking large SAN based media stores into
consideration
and then it is clear that careful setup of servers and storage in their racks can actually
improve performance and reduce power loads a lot mre than you would expect....
And, so I do not appear to be coming from this angle as a Jonesian fact maker,
here is some research to back it up!
http://www.dbms2.com/2010/05/08/disk-vibration-data-warehouse-performance-problem/
It is just as amazing to me, as when an engineer demonstrated SCSI reflections by
bending a cable beyond a certain limit, whereupon all data transmission ceased. Â The waveform
of the electonic stream had a period that meant it could not "bend" around the pathway made by the cable and the data just stopped moving. Â Â Â I find things like this amazing because it is like
observing some kinds of weird "quantum" behaviour you would not expect, manifesting
in the physical world.
rachel
--
Rachel Polanskis Kingswood, Greater Western Sydney, Australia
[email protected] http://www.zeta.org.au/~grove/grove.html
"The perversity of the Universe tends towards a maximum." - Finagle's Law
--
SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/
Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
