At first glance, Ben Kirtman’s job seems like the stuff of science fiction. The University of Miami climatologist predicts the future of Earth’s climate in timescales of not only decades but also days.
Gary Beecham faces no less of a daunting challenge. To identify genes associated with certain diseases, the UM statistical geneticist must study the genetic variants in tens of thousands of individuals, analyzing as many as 200 billion data points.
Their research would take years to complete without powerful computers. Now, a new supercomputer recently unveiled by UM’s Center for Computational Science (CCS) is allowing Kirtman, Beecham, and other investigators across the University to perform complex calculations faster than ever before.
The new IBM-built Pegasus supercomputer came online this spring, offering UM researchers the ability to perform 160 trillion floating-point operations per second. Housed at the NAP of the Americas data center in Miami, it features a new generation of Intel central processing units, making it five times faster and more powerful than the previous Pegasus introduced by CCS about three years ago. Intel made the new state-of-the-art processing units, called Phi, available only to UM and a handful of other institutions.
“We outgrew the old Pegasus,” said CCS Director Nick Tsinoremas. “This new system is a quantum leap in computing power, and it comes with all the experience we’ve gained as a computing center over the past six years. Our goal has always been to have something this powerful.”
About 1,200 researchers used the old Pegasus, which remains operational and is now working in concert with its new ramped up big brother. Those users have migrated to the new Pegasus, accessing the system through remote login.
Kirtman and Beecham are among the newer system’s most prolific users. To determine what path Earth’s climate will take, Kirtman, who is a professor of meteorology and physical oceanography at UM’s Rosenstiel School of Marine and Atmospheric Science, constructs complex computer models that require a multitude of processors operating simultaneously. “The new Pegasus provides that capability,” he said.
With the completion of the Human Genome Project in 2003, the field of genetics has generated vast amounts of data, requiring faster and more efficient computers and computational methods to analyze such information. “The Pegasus cluster has been an answer to the big data challenge,” said Beecham, an assistant professor and director of research informatics at the John P. Hussman Institute for Human Genomics, a UM Miller School initiative that uses the latest technologies to identify genes involved in human diseases.
“Pegasus has allowed a level of parallelization that was previously infeasible with smaller computing systems,” said Beecham. “This lets us run these complex analyses at speeds that are on orders of magnitude faster than what was previously available.”
Kirtman and Beecham are just the tip of the iceberg in terms of the variety of investigators who will make use of the new Pegasus. Climatologists; economists; engineers; geneticists; atmospheric, oceanic, and, physical scientists; even musicians will use the system.
CCS is conducting training sessions to help investigators learn the system’s new bells and whistles, Tsinoremas said.
He added that the acquisition of the new Pegasus will make UM one of the top 500 supercomputing sites in the world and perhaps place it among the top 10 academic sites with the most powerful supercomputers.