PRObE NSF Facility

The NMC, the University of New Mexico (UNM), and Los Alamos National Laboratory (LANL) propose to create a NSF facility for low-level systems research. PRObE will accept decommissioned supercomputers from DOE facilities and make them available to the low-level systems computer science research community. Government and industry have long neglected the need for a large-scale research instrument in the low-level systems computer science community. Decommissioned very large-scale supercomputers provide an innovative way of addressing this need at a fraction of the cost of a classical applications-oriented supercomputer center. LANL and the NSF would support the Center jointly, with LANL providing the machines and some educational outreach and the NSF providing operations support and educational outreach.

NMC will house the facility as part of its operations in the Los Alamos Research Park. The Principal investigators are Gary Grider from LANL’s High Performance Computing (HPC) Division, and Tim Thomas and Andree Jacobson from UNM High Performance Computing Center. The facility would be scheduled for use by systems researchers funded by NSF or other government agencies, in blocks of days or weeks of time. The PRObE Center would either provide an open solicitation process modeled after other NSF large research instruments or utilize the NSF solicitation and review process itself for allowing users to use the facility.

The facility will also support a summer school in clusters and networking for New Mexico students.

The first computer will consist of portions of Lightning and Bolt (3570 node dual proc single and dual core opterons systems). LANL will de-commission these machines in the Fall of FY ’10, which is also the target date for completing the facility and beginning operations. The facility will require modifications to space in the Research Park to create a 2500 sq feet computer room. It will also take 3-5 offices in the existing NMC facility. The facility will employ ambient air cooling and will be a resource for the local computing community in the implementation of this method. This method can reduce the power required to cool the machine by more than 80%. The expected cost of the facility over five years is about $8M, approximately equally split among infrastructure improvements, staff and operating costs. The program supports a staff of four plus part-time positions for the Principal Investigator (PI) and summer school faculty.

Users will work under agreements that return Intellectual Property (IP) rights to their home institution. The researcher’s home institution will have liability for researchers while they work on the machine and they will be responsible for enforcing export control laws for the work they support at the facility. The grant will support increased general liability insurance and property insurance as appropriate for the NMC to cover visiting crews.

Contact: Katharine Chartrand knc@newmexicoconsortium.org