My rough notes from the first two sessions at the Data Center Efficiency Summit at Google Mountain view earlier today:
Data Center Energy Going Forward
· Speaker: John Tuccillo, APC
· Green Grid:
o Data Collection & Analysis
o Data Center Technology & Strategy
o Data Center Operations
o Data Center Metrics & Measurements
· Metrics team:
o PUE & DCiE
o DCP: Data Center Productivity
Insights in Google’s PUE Results
· Speakers: Chris Malone & Ben Jai, Google
· Chris started off by reviewed existing data from 6 data center average quarterly and published for a year (on web):
o All less than 1.3
o Best at 1.16 (Google DC ‘E’)
· Inclusion in external published data:
o 5MW or bigger and operating for more than 6 months
· Typical PUE ~1.7
· Google DC E
o Mechanical: (didn’t get data point)>
o Power Distribution: 4.9%
· Achieved by rigorous application of best practices:
o Air-side economization
o Water-side economization
o Close coupled cooling
o 99.9% UPS efficiency
· 99.9% UPS Efficiency (Ben Jai presenting)
o Distributed on-board UPS
o Single voltage motherboard (12v)
o Motherboard provides 5v to disk and all step downs needed by on board requirements
o Installed a lead-acid distributed UPS to ride through power sags
o Avoids double conversion of many central UPS
o Only enough power in UPS to allow generators to start or to switch to other A/C supply
· Google Measurement of PUE (Chris Malone):
o Average DC around PUE of 2.0 in 2006
o Sate of the art data center around 1.2 using exotic techniques
o 2 of 6 DC report daily, 4 of 6 report continuously
o Measure at sub-station and extrapolate to utility input at substation
o Most measurements on the server side taken at PDUs. On newer servers, it’s measured at PDUs (more precise).
o Accuracy of PUE measurement at +/-2%
· Best Google facility on quarterly basis: PUE => 1.19
o The problem with non-annual numbers is they are skewed by the impacts of changing weather conditions. Need to annualize to gain full insight.
o They should some impacts on PUE of weather factors and DC maintenance
o Showed utilization at different facilities:
§ Ranged from clusters around 30% to clusters on the high end at 75% (amazingly high by industry standards).
Chris and Ben presented great material in this last section. Super interesting, very nice designs, and well presented. The PUE measurement techniques look credible and the results are excellent.
–jrh
James Hamilton, Amazon Web Services
1200, 12th Ave. S., Seattle, WA, 98144
W:+1(425)703-9972 | C:+1(206)910-4692 | H:+1(206)201-1859 | james@amazon.com
H:mvdirona.com | W:mvdirona.com/jrh/work | blog:http://perspectives.mvdirona.com
Abe, I’m interested in what you folks at Ballard have suitable for application in a data center. If you have anything you can send my way, please do. Thanks
–jrh
jrh@mvdirona.com
Today’s Google announcement on Back up with independent batteries mounted with every server is getting thumbs up.
You still need batteries that need charging, monitoring and replacing. Consumption and supply chain is not eliminated.
Having a Fuel cell Back-Up system in a data container that only turns on when needed should be a valid value propositions.