Breaking Open the Unknown Universe

Vintage Store: The quickest, cheapest, most reliable way to store all that data? Tape drives, same as in the 1970s:  Maximilien Brice/CERN

The Data Junkies

Back in the cavern that holds Atlas, physicist/tour guide Steve Goldfarb stands on a gantry 50 feet above the floor and traces in the air an imaginary track of an imaginary particle that has just spawned from a collision. "The whole idea of building such a huge detector," he says, "is to be able to draw a very precise line." Tellingly, the line he draws curves across the room.

Both Atlas and CMS generate magnetic fields so intense that "if you drove a bus in here and if you turned on the magnetic field, you would crush the bus," says Phil Harris, a graduate student at the Massachusetts Institute of Technology who shows me around CMS the following day. (Graduate students are considered the do-it-all grunt workers of any enormous project like this. Harris's buddy Pieter Everaerts, another MIT grad student, told me that one of their main jobs was to "go down [into the detector] to look for the blinking lights" that may indicate a faulty connection. Harris, for his part, has spent months building a database to keep track of the thousands of cables that carry data up and out of the machine. The LHC: where America's best and brightest go to label cables.)

Bus-crushing, despite its indisputable awesomeness, is not on the agenda here. Rather, the point of all these superconducting magnets is to make everything curve. When the two protons collide, the shower of debris they create will not, unlike the cables in the detector, come with labels. Harris and Everaerts and the 2,000 other scientists who work on CMS have to figure out what each particle is. Since a magnetic field bends the path of a charged particle, you can measure how much each particle curves and how fast it's going and deduce its charge and mass. "We need to understand everything," Harris explains. "Where it was, how much momentum, how much energy." And do it over and over, for the hundreds of particles that burst from every collision, 600 million times a second.

This, in turn, presents a slight problem with data overload. "We'll produce about a World Wide Web's worth of data every day," says Harris, an excitable 25-year-old who wears his hard hat backward and his pants a good six to eight inches below his waist. Everaerts turns his eyes up, clearly checking the math behind Harris's boast in his head. "Yes," he solemnly intones, "though the Web is growing very fast."

It's one thing to undertake a massive (but finite) civil-engineering project like the LHC in the space of a decade. It's quite another to build a new Google every day. "There's no way that CERN can provide all the computing components," says Ian Bird, the leader of the LHC Computing Grid. Instead, scientists figured out two ways to get rid of all the excess data.

Fortunately (or not, depending on how you look at it), most of the data the machines collect will be junk. Old news, particles long discovered, phenomena well-explored. Electronics in the detector throw out any collisions that don't look interesting, which totals about 99.99997 percent of the raw data.

The remaining 200 collisions per second move upstairs to the main computing center, a warehouse with row after row of rack-mounted computers. This is "Tier 0," in LHC parlance. From here, dedicated fiber-optic cables send a copy of the data to 11 computing centers worldwide, the so-called Tier 1. (The cables comprise the famous "Internet2" you may have heard about a few years ago—all it means is that the scientists get to use these lines, not you.) The Tier 1 computers then calibrate the data and distribute it to hundreds of Tier 2 computing centers. These are individual server farms, the 100,000 PCs spread among universities like Cambridge and Berkeley and Osaka. This is where the eureka moments will happen. By using a distributed system, the collisions underneath a French village can branch out all over the planet to be pored over by 10,000 brains. It is through this structure, just as much as through the magnets or the silicon, that the impossible will be made real.