\n The Parallel Reconfigurable Observational Environment (PRObE) was a seven year long project, funded by the National Science Foundation. It set out to address the question of whether large scale government computer computer systems, slated for retirement, can be repurposed by a non-government entity and again used for disruptive research. With today’s computer systems’ rapidly increasing processing speeds, it is clear that such a resource can not be cost effectively used for scientific simulations, as that is the reason they are retired after only three to four years in the first place. However, there are research areas where experiments on these types of large systems are critical to progress in the field. Low-level (close to the hardware) computer systems research is the perfect match. Disruptive experiments, such that would hamper the work of others on the same resource (rebooting a computer for example), are often required in this type of research. PRObE provided exactly this type of system, with some clusters consisting of more than 1,000 computers. For five years, the resource was available for researchers that laid the ground to many technologies that today drive what is popularly called “the cloud”, or cloud computing. Research areas specifically helped by this effort included operating systems, networking, file systems, resilient computing, and others. The experimental PRObE facility at the New Mexico Consortium in Los Alamos, New Mexico closed its doors for services on Dec 31, 2016. The project officially concluded on June 30th, 2017. During this time, PRObE received and repurposed thousands of computers from Los Alamos National Laboratory in a unique center without equal elsewhere in the world. PRObE, at peak capacity, provided around 1,500 fully operational computers in four clusters in New Mexico and at Carnegie Mellon University. The University of Utah was also a major collaborative partner. Research performed at the center resulted in 44 peer reviewed conference proceedings, 5 published journal articles, and numerous workshops, presentations, white-papers, technical reports, and graduate dissertations. Over 500 researchers requested access to the resources, and PRObE provided around 170,000,000 core hours of computation. Purchasing something similar from current cloud providers might cost on the order of $15M. Overall, the PRObE project provided a fantastic resource for a research community that did not have any testbeds at this scale. It proved that it is possible to forge collaborations between government, academia, and nonprofit organizations in a research community. Many researchers were able to perform research they would not otherwise have been able to do, and solid support was provided by the community throughout the project. The New Mexico Consortium is honored to have been part of this valuable project, and particularly the profound impact it had on research.
\n\t\t\t\t \tLast Modified: 09/29/2017
\n\t\t\t\t\tModified by: Gary Grider
PRObE is the Parallel Reconfigurable Observational Environment - An NSF funded low-level large-scale computer systems research facility located in Los Alamos, NM. Housed by the New Mexico Consortium (NMC) at the Los Alamos Research Park (LARP) across the street from Los Alamos National Laboratory (LANL), PRObE’s main mission is to provide low-level access to several large scale testbeds (1000+ nodes) built from repurposed supercomputers from LANL.
\nPRObE is a collaborative project whose key partners are NMC, LANL, Carnegie Mellon University (CMU), and University of Utah (UofU).
\nAs computer systems grow and become more powerful, the world's largest scale research computers are reserved for domain science computations. Little time on such expensive computer clusters is dedicated to understanding the large scale computer systems themselves in order to improve these systems or to train the workforce for these large computers. PRObE, however, is a unique facility that provides exactly this type of testbed. Its target audience includes computer systems researchers studying operating systems, scalable architectures, file systems and storage technologies, complex network architectures, cloud deployment, and similar topics. To our knowledge, PRObE is the only facility of it’s kind in the world.
\nThe key outcome for the PRObE project was the successful repurposing to NSF use of retired supercomputers funded initially by Department of Energy activities and operated at Los Alamos National Laboratory. This equipment can be useful for computer systems research for years after it is inappropriate for its original purposes. We recommend that this opportunity, repurposing retired national computing resources for NSF research, be further pursued, and recommend that the New Mexico Consortium facilities would be appropriate for cost-effective capture of large amounts of Los Alamos National Laboratory retired computers.
\nThe second key outcome for the PRObE project was exploitation of repurposed supercomputers for low-level computer systems research. In only a few years of operation, with no existing base of users, over 50 institutions requested and used the clusters, with almost 400 users. Universities such as Cornell, Georgia Tech, Harvard, Illinois Inst of Tech, Indiana, MIT, NYU, Penn State, Princeton, Purdue, UC Berkeley, UC San Diego, Central Florida, Chicago, New Mexico, Pittsburgh, USC, UT Austin, Utah, Virginia, Wisconsin at Madison, and Virginia Tech used the facilities in their computer systems research, in addition to labs such as Argonne, Los Alamos, NCSA and Oak Ridge.
\n\t\t\t\t \tLast Modified: 07/13/2016
\n\t\t\t\t\tModified by: Garth A Gibson