After its tenure at Los Alamos National Laboratory (LANL), the Ulam supercomputer hit its retirement age in early 2014. But rather than simply decommissioning the system, the New Mexico Consortium opted to let it live out the rest of its days serving another community: the University of New Mexico (UNM).
Now housed in what was once a car dealership along Route 66, Ulam (named for Polish mathematician, Manhattan Project member and inventor of the Monte Carlo computation method, Stanislaw Ulam) is able to offer students the same compute power reserved for the nation’s top national labs.
The transition began roughly a year ago with one shipment containing 120 nodes, followed by two more that delivered the racks, switches and cables. But now as UNM finishes renovating the system’s machine room to include a 30-ton air conditioning unit, an uninterruptable power supply and backup power system and hot aisle containment for energy efficiency, the system is ready to go online.
In addition to commissioning the 13-rack system, UNM’s expanded machine room will also house Xena, a GPU-accelerated cluster from the SNF’s Major Research Instrumentation Program (MRI), as well as the existing Metropolis system that also came from LANL through the New Mexico Consortium. Nonetheless, Ulam is expected to fill the university’s top computing spot.
“There is more memory on each node of Ulam than on most of our other machines,” said Susan Atlas, director of the UNM Center for Adanced Research Computing. “There are also more nodes and nearly 1,000 cores so researchers can tackle significantly larger problems at our center than previously possible.”
The addition of Ulam will raise the center’s total capacity to approximately 3,000 cores.
Both Ulam and Metropolis were supplied by the New Mexico Consortium through the NSF-funded PRObE project, whose mission is to supply decommissioned systems from Los Alamos to institutions such as UNM.
The university predicts that most of Ulam’s new users will come from the school’s Biology Department and Cancer Center to power genomic and bioinformatics research, while also offering machine learning and pattern recognition abilities for computational biology, neuroscience, computer science and astrophysics.