Video: Google’s Data Center Best Practices
The primary goal of the Google European Data Center Summit held Tuesday in Zurich is to advance the adoption of well-known best practices to help make data centers more efficient. The event showcased some case studies from Google, and also featured many industry experts discussing various best practices that can reduce energy usage. The collective recommendations and best practices have been compiled in this video, narrated by Google Technical Program Manager Erik Teetzel and featuring other members of the Google data center team. The video runs 10 minutes, and provides an executive overview of data center best practices.
Johnny MnemonicPosted May 25th, 2011
Article is bs. If you watch closely, you’ll notice that they only discuss what is done in the CCNR and corporate DC space. That accounts for perhaps 5% of Google’s total DC floorspace. Therefore the “best practices” as described here should not be confused with what Google does in their production environment.
There’s a nice point made about the UPS, but wouldn’t putting a battery on each machine lead to a management and maintenance problem when those batteries begin to fail at their MTBF? What do you do with thousands of leaking batteries?
Nuno GarciaPosted May 26th, 2011
Comment to Johnny Mnemonic:
1. it is very immature to say that the article is **. Please grow up.
2. it is true, the article does not discuss the overall issue of corporate DC space. Nevertheless, the article is well presented and the rules that are suggested are sound and important, even if applicable only to 5% of the company’s DC space.
3. The MTBF for batteries is more or less the same as the MTB for the other components, so you end up having a whole machine that has to be replaced. It makes a lot of sense to eliminate unnecessary power conversions, also rendering the solution a lot more scalable – UPS solutions are not easily scalable.
JeffPosted May 26th, 2011
For Google’s typical use, a server will be outdated and obsolete before the battery hits its MTBF (not counting premature failure) so why worry about that? As long as the server is a sealed box, if it develops a bad battery or something they will just pull it out and throw it away. The batteries will live longer than the CPU will be useful for, and (hopefully) they will just ship obsolete servers by the container-full to China for recycling. Google has a lot of impressive innovation but their methodologies are all centered around cloud computing approaches where servers are cheap, expendable, and fast to expire.
This all means that if you want to run your data center like Google, you need to run your applications like Google; something that’s really not an option for most companies, even large ones.