Computers are getting smaller, hard drives are getting cheaper, and processors are getting faster — but that doesn't mean we no longer have computers so huge that they fill entire floors of a building. It just means that when computers do fill an entire floor, they are crunching a shitload more data than they would have back in the mainframe days. These days, building-sized datacenters serve up everything from highly-granular genomic data to web pages that were deleted five years ago. Taking you deep inside these datacenters, full of massive, icy server rooms and crazed sysadmins, is scifi writer and digital rights activist Cory Doctorow. He's written a great article for Nature about visiting some of the biggest datacenters in the world, and it's free online. Here is Doctorow at CERN, the Swiss research center where Tim Berners-Lee invented the World Wide Web and where the Large Hadron Collider is about to recreate a tiny version of the Big Bang:
The Vader-black machines, one built by StorageTek, a subsidiary of Sun Microsystems, the other by IBM, are housed in square, meshed-in casings the size of small shipping containers. From within them comes a continuous clacking noise like the rattling of steel polyhedral dice on a giant's Dungeons & Dragons table. I pressed my face against the mesh and peered in fascination at the robot arms zipping back and forth with tiny, precise movements, loading and unloading 500-GB tapes with the serene grace of Shaolin monks. Did I say tape is tetchy? I take it back. Tape is beautiful.
The best part of this article is that it's not about the fancy companies that own these machines, and it's not about the fancy things they're doing with their fancy data. Instead, the article takes the perspective of sysadmins, the people who are actually down in the trenches maintaining the machines:
At each data centre I asked the sysadmins for their worst fears. Universally, the answer was heat. Data centres are laid out in alternating cool and hot aisles, the cool looking at the front of the racks, the hot at the back. At CERN, they actually glass over the cool aisles to lower the cooling requirements, turning them into thrumming walk-in fridges lined with millions of tiny, twinkling lights.
This article isn't just about how petabyte machines provide elegant solutions to hard problems. It's about tending the machines themselves, maintaining the data infrastructure that we so often forget about. This is one of the best hands-on techie stories I've read in a while. Check it out! Big Data: Welcome to the Petacentre [via Nature]