Google's Datacenters Come to Light
A couple of sites have put together some interesting information on Google's clandestine data centers.
Data Center Knowledge has logged an incredible amount of data about the 36 or so centers worldwide, painting a broad picture of what to many users were almost mythical entities. Royal Pingdom followed this up by mapping out the locations for us.
The post lists the locations of the sites, including some under construction, such as one in Goose Creek, S.C., as well as Pryor, Oklahoma, Council Bluffs, Iowa—real heartland stuff after building data centers in more urban areas such as in its Mountain View hometown and Los Angeles.
Many know that Google keeps mum on their data centers for the same reason we don't know exactly how their search engine works—to keep the competition off balance.
What most probably don't know is that the company takes such steps to obfuscate data center details by getting permits for its data center projects using LLCs (limited liability corporations) that don't mention Google, such as Lapis LLC in North Carolina and Tetra LLC in Iowa. Also, you can't really find the data centers on Google's own search engine because the IP addresses point back to Mountain View.
So, what's a person thirsty for Google data center information to do? Short of cutting through or pole-vaulting over the fences at these massive complexes and risking arrest to poke around, you can read the second and third pages of the Data Center Knowledge post, where you'll learn about square footage, approximate cost for one data center ($600 million!), power consumed and gear employed.
The last is the most interesting detail to me. Data Center Knowledge repeats the mantra that Google uses commodity Web servers. These are custom built, include Intel and AMD processors and are typically stacked in great racks.
Google employs one to two processor machines running various iterations of the Linux operating system and Google infrastructure software, including Google File System, Google Web Server, Google MapReduce for parallel programming and Google BigTable for data storage.
Why am I so interested in all of this? I plan to attend the opening of one of these institutions of higher computing this year. I have questions: What are these places like in the metal? What do the residents think about them? What do the creepily titled transient employees hired to tend to them think about living at these data centers?
I'll keep you posted on GoogleWatch.