[AusNOG] Data Suburb

[Lincoln Dale]

On Thu, Jan 5, 2012 at 8:00 AM, Mark Newton <newton at atdot.dotat.org> wrote:

> I'm not sure that 1.24 counts as "poor." Especially when it's inside the
> ballpark of the contrary example you've cited :)
>

its also not "poor" given the ambient temperatures seen in the geography
where the site is located.
the physical site is more about where there is "cheap power" and "cheap
land", not "optimal ambient temperatures".
i'd hate to see how much water they go through though!

comparison of PUE is also meaningless when it comes to different
application profiles.  nothing against Phillip but comparing GOOG's
workload to that of a generic colo is apples/oranges comparison. comparing
Switch to Amazon AWS or EC2 is more meaningful.
(James Hamilton @ Amazon has posted their data (
http://www.datacenterknowledge.com/archives/2011/06/09/a-look-inside-amazons-data-centers/)
which in their case shows a cost breakdown, and even for a PUE of 1.45
shows majority of cost is NOT in the cooling.)

comparing GOOG's PUE should be done with Facebook's Prineville facility or
FB's Swedish facility under construction.
or comparing what the plans are with Open Compute servers.


cheers,

lincoln.

-- 

Lincoln Dale | Principle Engineer | Arista Networks
ltd at aristanetworks.com | +1 408 547 5782  | +61 417 457 965
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ausnog.net/pipermail/ausnog/attachments/20120105/4ab5c8ce/attachment.html>