In anticipation of the enormous amounts of data that will pour out of the LHC, researchers have turned to grid computing, a new way of computing that distributes resources and data among institutions around the world. But to run this vast and complex network of distributed computing, researchers need sophisticated software to match the task.
"This is the first time SLAC, the high energy physics community, and even the world are trying grid computing as the only computing model to support an experiment," said Wei Yang, a SLAC software engineer in the ATLAS collaboration. "It's a difficult task, and SLAC is collaborating with institutions around the world to make it work."
Once the LHC is up and running, computing systems will not only need to handle massive amounts of generated data, but will also have to account for simulation and analysis requests from around the world. To make this possible, LHC researchers, including those at SLAC, are working on software that shuttles tasks and data among dozens of computing facilities worldwide. This software allows researchers to apply computer science theories to real scientific problems.
"Computer scientists make all kinds of models and theories, but an experiment like ATLAS will determine which one actually works," said Randy Melen, head of the Scientific Computing and Computing Services high performance computing team.
Because of the sheer scale of the project, the software has a large number of interconnected components, making it an ongoing challenge to ensure smooth operation. In addition, SLAC software engineers have to keep up with security fixes.
"It's a challenge, but grid-based LHC software will make this distributed environment work," Yang said. The underlying technology and software will not only enable efficient LHC data analysis, but will be used for all types of scientific computing, including biology and chemistry.
óMarcus Woo, March 30, 2007
Above image, from left to right: Wei Yang, John Bartelt, Leonard Moss, Randy Melen, Charles Boeheim, and Booker Bense.