September 10th, 2008
The Collider and the Grid: Distributed computing made LHC possible
So the LHC is turned on! And it works great. And we’re still alive. I’m not gonna beat that horse today. (See Ars for snark)
But a good IT angle here. LHC is really only possible because of the growth of grid computing. This quote from Computerworld says it well:
“The distributed computing model is essential to doing the computing, storage and hosting of the many petabytes of data from the experiments,” said [Harvey] Newman, [professor at California Institute of Technology]. “Coordinating data distribution, processing and analysis of the data collaboratively by a worldwide community of scientists working on the LHC are key to the physics discoveries. Only a worldwide effort could provide the resources needed.”
Charles King of Pund-IT points me to several IBM press releases over the year regarding their support for LHC. For instance, here’s a release from last year about an IBM supercomputer being used at Canada’s national laboratory for particle and nuclear physics.
After passing extensive acceptance tests, the system has been integrated into an international “grid” of high-performance computing centres that will analyze data from the global ATLAS project, commencing in the summer of 2008.
Back in 2004, Wired had a piece on the LHC’s grid network. Les Robertson, who studied with Peter Higgs (of the boson) and head of the grid, realized distributed computing was the way to go.
“I had assumed we’d do this in a simple way,” he says. “We’d develop the management software ourselves and pass bits of data down an inverted tree” - in other words, data would flow to nearby centers, and results would come back in a hierarchical, predictable way. The centers wouldn’t be linked into any kind of virtual supercomputer; they wouldn’t have to be.”
