geek job #11: Computer Jock for the LHC
The teams I’ve worked on have managed some incredibly large data centers. But I have some serious geek envy for the dudes running the computing and network infrastructure for CERN’s Large Haldron Collider.
In case you’ve been sleeping in a scientific vacuum, you know that that the Large Haldron Collider (LHC) is where big bucks meets particle physics. To says that it is “big” is using small words for, well, big things:
- it is the most expensive science experiment, with a budget of close to $6 billion; the cost is so massive that it is almost an all or nothing bet – many countries have put such a large percentage of their science budget into the LHC that there may not be funds for other experiments until past 2010;
- it is physically one of the largest; heck it’s in Switzerland and France;
- its cryogenic system could easily cool the warm beer server up the road in Munich at Oktoberfest (i.e., it is the world’s largest refrigerator);
- two beams of protons traveling at 99.99% of the speed of light, will generate an energy of 7 TeV (terra electron volts), with the collision equally 14Tev.
- it will utilize massive computing power and storage; detection arrays will see close to a 600 million collisions per second; the compute infrastructure must determine (quickly!) which of those collisions is worth watching/researching, and then collect the large amounts of data needed. Estimates for the total storage requirements for the LHC run close to 10 petabytes per year (that’s 10 million gigabytes).
The people managing the wealth of compute and network horsepower for this project are in pure geek heaven. Because of how fast the action is occuring and how quickly decisions need to be made on it, there is some serious CPU’s at the frontline. And the data has to be stored and distributed, so there’s some nice networks and storage to play with.
Near as I can tell (and anyone with more knowledge please leave some geek appropriate comments) there are six levels of compute infrastructure:
- the initial detectors, built in layers to detect certain types of particles;
- a computer-board trigger system, which makes decisions about what detections are worth looking at;
- a higher-level trigger system, made up of tons of computers, which will rebuild the collision based on the data;
- the LHC computing Grid, a big ole farm of computers (most of them, as I understand, not bleeding edge but blade systems purchased at the best price/performance level (can you say volume purchase agreement?);
- 12 data centers around the world connected by hi-speed optical lines, so the data can be studied in near-real time at places like Fermi and Brookhaven;
- universities that supply more gird power.
The highest profile experiment is the search and detection of the Higgs boson. Predicted in the 1960’s, the theory to be tested is that everywhere there are Higgs fields, and particles that interact within these fields are given mass via the Higgs boson. This theory is a required part of the theoretical mathematics that unified two of the four fundamental forces, electromagnetism and the weak nuclear force.
Detection of the Higgs boson will be difficult, as its mass is not known. But the find would solve the mystery of why particles have certain mass; not finding or detecting it would open up the concept of particle mass to the world of physics theory.
Another experiment, known as ALICE (A Large Ion Collider Experiment) aims to allow physicists to study matter as it was immediately after the beginning of the universe (right after the big bang). The heat of the collisions at the LHC should produce a state of matter know as quark-gluon plasma; scientists will then observe this state of matter, hoping to understand what happened in the moments after the big bang.
The LHC may also produce tiny black holes, which would deteriorate very quickly, giving off a particular kind of radiation that the experiments could detect and collect.
Another possible production of the LHC would be new as yet unpredicted particles.