Berkeley Lab and the Stanford Linear Accelerator Center have a
long history of collaboration stemming from their high energy physics research. Now the
two facilities are leveraging their resources in the cutting edge area of scientific
computing.
Among the NERSC-SLAC collaborations in high energy
and nuclear physics is the experiment known as BaBar. BaBar's magnet is pictured here.
|
Examples range from sharing experiences with software applications to developing new
systems for gathering, organizing and analyzing data from the upcoming series of particle
experiments known as BaBar.
Enhancing scientific computing among DOE labs was one of the objectives of the 1996
move of the Energy Sciences Network and the National Energy Research Scientific Computing
Center (NERSC) to Berkeley Lab. Richard Mount, director of SLAC Computing Services, says
the move is already bridging what had been a big gap between the scientific computing
community and the scientific data processing fraternity within high energy and nuclear
physics.
The proximity of the two DOE facilities, combined with the convergence of scientific
computing and scientific data processing, are providing new avenues for cooperation.
"There has been a real effort by NERSC to move into scientific computing and
scientific data management," Mount said. "There is also a strong direction in
high energy physics toward increasingly data intensive experiments. We're seeing a strong
effort to seek out common solutions, to seek out new possibilities."
One focal point of this effort is DOE's High Energy and Nuclear Physics Data Grand
Challenge to develop tools to allow high energy and nuclear physicists to analyze and
manage the massive amounts of data from future experiments, such as those planned for the
Relativistic Heavy Ion Collider at Brookhaven National Lab. These experiments, aimed at
understanding quarks and other exotic forms of matter formed in the wake of the Big Bang,
will generate huge amounts of data--hundreds of terabytes a year.
A large part of the challenge in the early stages will be to predict data-access and
computational problems and find ways to solve them. NERSC and SLAC are both working on the
problem.
The Computing Sciences organization at Berkeley Lab has several groups working on
scientific computing solutions to high energy physics problems. These include the
Visualization, Distributed Collaboration, HENP Support, Scientific Computing, Scientific
Data Management, and HEP Computing groups.
"The scientific computing problems challenging high energy physics are also
showing up in other fields of research," says NERSC director Horst Simon.
"Genomics researchers also must find ways to organize and access vast banks of data.
Computing Sciences is working on solutions covering many aspects of research, from
gathering data from experiments, to organizing, analyzing and presenting the data in
useful ways."
For example, the six-person HENP Support Group in the Lab's Information and Computing
Sciences Division is designing computer tools to help with the BaBar experiments to be
conducted at the "B Factory" being built at SLAC. The goal is to measure the
decays of subatomic particles and anti-particles known as mesons. To acquire a sufficient
sample of those decays that can reveal fundamental differences between matter and
anti-matter, researchers will need to perform up to one billion experiments a year,
yielding 150 terabytes of data. They will then need powerful statistical analysis tools to
analyze the information.
"This is the most advanced software project in high energy nuclear physics
today," said Information and Computing Sciences Division leader Stu Loken.
"Ex-periments at future particle detectors will depend on the work we're doing for
BaBar."
About 500 physicists in 10 countries are participating in this project and need access
to the data, which will be stored in NERSC's High Performance Storage Systems. "We
really feel we are welcome collaborators," Mount said.
This situation is a turnaround from the early 1980s, when high energy physics
experiment data were recorded on magnetic tapes, and it was not uncommon to fill 10,000
tapes with data. Showing up at a supercomputing center with the equivalent of two
truckloads of tape to process did not produce a warm reception at a time when the machines
were less powerful, with limited tape handling facilities, Mount said.
To enhance the collaboration between the two facilities, SLAC now provides office space
to NERSC employees who live in the South Bay and Peninsula and choose to telecommute.
Employees from both facilities routinely work at each other's sites, applying their
expertise and drawing upon that of their counterparts. And Berkeley Lab's Stu Loken and
Carl Eben, deputy division leader of ICSD, serve on SLAC's computing review committee.
"Our respective resources put us in a unique position to contribute to the
Department of Energy and the nation as a whole," Mount said. "I see this growing
cooperation as a strategic effort, not just a technical sideline."