Inside Facebook’s new data center: http://blog.facebook.com/blog.php?post=262655797130
My thoughts on Facebook’s new data center:
The BULK of what Facebook has done at this facility follows what we’d all define as best practices; buy the most efficient gear for your apps – measuring and taking into account transactions per watt (if you can get the manufacturer to do a custom sku – more power to you), use the most efficient cooling for your environment (ASHRAE’s recommendation vs what server manufacturers will warrantee as inlet temperature and humidity envelopes increase is leaning towards ASHRAE updating their recommendations to what manufacturers are comfortable with – see HP for proof), have a clearly defined thermal containment strategy, find creative ways to reuse the waste product of your facility (water, air).
I agree completely with Facebook’s approach of building application redundant environments, rather than overbuilding redundancy inside facilities. I think we can all agree there. When the application is designed to be agile and highly portable – everyone benefits. If every enterprise that was considering building their next Tier IV data center, instead – first considered building more reliable and more agile applications (when latency appropriate), we’d live in a World with more efficient, more reliable applications that result in more efficient and more reliable facilities by the merit of their design.
Regardless of the margin of error (which I doubt is a ton) in measuring and reporting their PUE; I’m sure these guys are thinking more along the lines of CADE or something meaningful to help them plan their insane growth. I, for one, am looking forward to CADE outpacing PUE as valued metric of efficiency in our data centers; offering data center managers and IT leaders real intelligence around their facilities and applications.
-Alex Fielding (FDCCI)