A new high-performance computing resource previously only accessible for campus researchers from its location at Idaho National Laboratory’s data center in Idaho Falls is now available on campus with a large visualization wall. Faculty may request an account or schedule use of the visualization wall on the Kestrel cluster by emailing Inanc Senocak at firstname.lastname@example.org or Jason Cook at email@example.com.
The Kestrel cluster attacks your Big Data problems with a fast 64 TB parallel file storage system for large data sets, 2TB of memory, 512 Intel Xeon 2.6 GHz computing cores, and 64 high end graphics processing units (GPU).
Acquired with a grant from the National Science Foundation’s Major Research Instrumentation Program, Kestrel is capable of working through complex problems across disciplines, including projects as diverse as wind energy forecasting, modeling for threat reduction in chemical and biological defense, materials characterization and modeling, snow hydrology and remote sensing, and mechanisms of skeleton development in living systems.
“Kestrel’s software environment replicates the environment that is available on national supercomputers,” said principal investigator Senocak, associate professor of mechanical and biomedical engineering. “Kestrel may not be a Top 500 supercomputer, but it sure provides a similar hardware and software environment to campus researchers at a modest size.”
The system includes a visualization wall comprising 40 connected 30-inch high-end monitors, meaning images can be viewed in incredible detail and clarity at up to 100 megapixels. Or, if preferred, researchers can visualize their data remotely on their own office computer monitors through the client-server mode with all the rendering done on the Kestrel cluster.
Built in-house, the viz-wall uses open-source software. A variety of software programs also are available on Kestrel.
“We will work with faculty to have the software they need on Kestrel,” said Jason Cook, systems administrator for the College of Engineering who spent long hours custom building several open-source software options to run efficiently on Kestrel.
And because the job is not finished after making the software available, Cook has ensured performance and scaling across many nodes of the cluster. Current research projects at Senocak’s research lab provide Cook with the simulation data to stress-test the software environment up to the hardware limit.
Kestrel cluster is making a difference in day-to-day operations of researchers on campus.
Rey DeLeon, a graduate of Boise State’s Mechanical Engineering program (B.S. 2010, M.S. 2012) currently is a doctoral student at the University of Idaho, where he is continuing his research on complex terrain wind forecasting in Senocak’s research lab.
“Kestrel has enabled me to have access to GPU computing hardware and tackle wind simulations over five times larger than previous Boise State computing resources could have,” DeLeon said. “These problem sizes are significantly larger than most seen in my specific area of research. Not only that, Kestrel also helps me effectively and efficiently process and visualize the large amounts of data that are generated from these simulations.”
Eric Jankowski, who will join the Department of Materials Science and Engineering as an assistant professor in January 2015, already has started using Kestrel in his research to engineer new solar cells made from organic materials.
Jankowski uses molecular-level simulations to get organic molecules from ordered structures to improve conversion of light into electricity. Jankowski, who has been using the GPU cluster at the National Energy Research Scientific Computing Center for the past three years, said that moving to Kestrel has been a huge pleasure now that his simulations execute two to four times faster.
Compared to the GPU computing resources at the Texas Advanced Computing Center, Kestrel achieves nearly identical performance with better availability (something that helps Boise State users do more work, faster). Titan at the Oak Ridge National Laboratory, the second-fastest supercomputer in the world, is attractive to Jankowski as well, but their emphasis on large multi-node jobs hinders his ability to acquire research allocations on Titan.
Said Jankowski: “For the work my group is doing, there currently is no better computing cluster in the world than Kestrel.”
This project is based upon work supported by the National Science Foundation under Grant No. 1229709 to Boise State University. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.