Google clarifies PUE Post Critiques

I received a note from from Eitan Bencuya from Google hoping to clarify some of the concerns I raised regarding their recent information post.

An excerpt from the  note is below:

Tech Hermit -

I noticed your post on Google’s PUE announcement today, and just wanted to provide a little bit of background to help clarify some of the concerns you raised in your article.

I think there was some ambiguity over the phrase “Energy weighted annual PUE.” “Energy weighted” simply means that we weighed the facilities in proportion to their load, so the energy weighted PUE simply is the sum of all power across all facilities, divided by the sum of the IT equipment load of the facilities.  In other words it’s the true average PUE across these facilities.  (The unweighted average, which counts every data center the same, typically doesn’t provide an accurate representation, and could improperly weight a super-efficient data center with a low IT power). “Annual” means that it’s the annual average (except for the two facilities that only have data for 6 months.)  Many companies will quote the best case PUE taken in the winter, but we wanted to provide the average across the entire year.

Also, this data excludes leased facilities since the PUE data provided on the site is just for datacenters which we’ve built.  The point of our disclosure is to show what’s possible if you focus on efficiency from the ground up, and we host the majority of our load in our own facilities.

Finally, we interpret the PUE in the strictest form, i.e. using the smallest possible number for “IT Equipment Power”.  Overestimating the IT Equipment Power term by including distribution and line losses will yield an artificially low PUE value.  If you have any questions about the way PUE is calculated please let me know and I’d be happy to walk you through this in more detail.

Thanks and please do let me know if you have any further questions,

-Eitan

As I stated before, I applaud Google’s recent decision to share more on what they are doing.  I have long beleived that the big data center operators share more of the best practices and lead the industry.     Additionally I truly appreciate their desire to ensure that my concerns were addressed. 

I am still doubtful on a couple of points like how using evap cooling is good, but I am simply having a hard time with the 1.21 numbers.  Perhaps they dont have enough data points to get a truly accurate annual average. My point about posting their total average PUE would be more interesting than just the ones they are building.   Not everyone has the ability to build new facilities like Google, Microsoft, and Yahoo.  We have to deal with facilities we already own.   Understanding all of their other facilities that they don’t own or build (perhaps arguably la footprint arger than their owned facilities) and managing to a larger program would be more interesting to the industry at large.  My concerns remain however, I dont think they are going to let me see the actual calculations and data but its good to see some sharing at least.