According to Bloom, who is the project manager for tier-two grid computing in the U.S., grid computing was the only option available to the engineers making the CMS at the LHC a reality.
"Just the raw data coming off the detector will be petabytes per year, for which we have had to reinvent grid computing to handle," said Bloom. "There will be millions of collisions per second, yet we can only process that data at about 100 Hz, which means the EEs had to design a high-speed online system that was smart enough to record only the collisions of interest. Even so, it was not possible to get enough power and computing into a single data center to handle these massive data sets, so we have gone to a distributed model that is enabled by the fact that we have high-speed networks to move the data around the world."
The six experiments at the LHC, including the CMS, will produce about 15 petabytes of data per year, which will be recorded and stored at CERN, but will be analyzed by a worldwide grid computing network made accessible to over 5,000 scientists around the globe.
The Worldwide LHC Computing Grid is facilitated by the the Open Science Grid, which has more than 60 sites in the U.S. and five sites in Brazil, Taiwan and England. The Worldwide LHC Computing Grid consists of seven tier-one data centers (including Fermi National Accelerator Laboratory and Brookhaven National Laboratory) which will partition the data sets into groups appropriate for the six experiments to be performed by LHC scientists. Finally, 40 tier-two data centers (including seven in the U.S.) will allow scientists to apply analytic algorithms to the data sets appropriate to their particular experiments. The U.S. data centers will provide more than 10 petabytes of disk cache for simulation and analysis.
The tier-one data centers are all connected by 10-Gbit per second fiber optic links which will be utilized around the clock to stream and partition the data into appropritate sets. The tier-two data centers in the U.S. will make use of the Energy Sciences Network (ESnet) and the Internet2.
U.S. funding was provided by the National Science Foundation and the US Department of Energy Office of Science.