Miami University IT Services and Research Computing Support have been implementing a significant portion of the new high performance computing (HPC) cluster, named Redhawk3, over the last few months. The result: Redhawk3 went live for the Miami research community on October 11, at noon. While not at its full capacity yet, Redhawk3 already exceeds the compute power of the old system by far, in terms of available memory and compute speed.
The new machines have state-of-the art Intel processors that work at a clock speed of 2.6GHz, and each has 24 compute cores and about 100GB of shared memory. One large memory node has 1.5 TB of RAM. Now that the new system is being implemented, the old system will be decommissioned.
When users access the new cluster, they will find their home space and files from the old system transferred and readily available to them.
A new security feature will require users to authenticate with two-factor authorization, in line with the new Miami security policy. Current users of the HPC cluster can request detailed instructions on how to access the new cluster by emailing Research Computing Support.
By Jens Mueller, Senior Research Computing Specialist, Research Computing Support, Miami University.
HPC detail wiring photo by U.S. Department of Energy via Flickr, public domain. Photo of DJ Rao by Jeff Sabo, Miami University Photo Services.
Miami University is purchasing a new high performance computing (HPC) cluster this spring. It will serve Miami researchers — both faculty and students — who have extensive computational needs that typically cannot be met by personal or lab computers. Named Redhawk3, the cluster will be made up of machines with the latest Intel processor architecture as core components. Compute nodes will have 24 Intel Skylake Xeon Gold 6126 cores with a clock speed of 2.6GHz.
The new system will be much faster than the current Redhawk, with most computations completing several times faster. The compute nodes are equipped with 96GB of RAM, making it possible to target bigger and more challenging computational problems. Redhawk3 will have two large memory nodes, each with 1.5TB of RAM. These nodes are meant to support researchers who have an interest in big data analytics or machine learning. In addition, the new cluster will have several nodes with general purpose graphics processing units (GPGPUs). GPUs have become increasingly popular because they can significantly speed up computations in a wide array of disciplines, including bioinformatics, engineering, physics, chemistry and many more.
Redhawk3 will facilitate the work of researchers like Scott Hartley, professor of chemistry and biochemistry; DJ Rao, assistant professor of computer science and software engineering; and Nam Vu, assistant professor of economics.
One of the key projects in Hartley’s group focuses on understanding how molecules fold, and how that folding can be controlled to give predictable structural complexity. Hartley says computational chemistry calculations — which demand a lot of computing power — are an invaluable resource for this sort of work.
“They let us predict behavior before embarking on time-consuming synthesis, and then allow us to relate our experimental data to molecular structure,” Hartley says. “As our understanding of our system has increased, so have our computational needs. The new cluster will allow us to explore ever-more sophisticated molecular architectures.”
Rao and his team of undergraduate and graduate students are developing a novel computational methodology for predicting and forecasting emergent epidemics in humans and livestock, including Zika, chikungunya, dengue, and avian influenza. The computational method focuses on parallel simulations and machine learning using detailed mechanistic models that simulate the lifecycle of almost a billion humans in 51 countries in the Americas, along with mosquito lifecycles, weather, and air traffic. Rao says such a large and complex model would require over three months of run-time on a conventional desktop, making it impractical. However, with supercomputing, processing time is reduced to under 24 hours, enabling global health research that is of national and international importance.
Vu’s research focuses on understanding the extent to which fiscal and monetary policy can impact the economy. Like Rao’s, Vu’s work typically requires computations that would take multiple days to process, even on a very powerful desktop computer.
“The old Redhawk cluster has been critical in my research agenda in terms of computing time and precision,” he says. “I have no doubt that the new Redhawk3 cluster, along with the wonderful support provided by the Research Computing Support group, will benefit my research for years to come.”
Redhawk3 is projected to be fully operational by the beginning of the fall semester, after an initial testing period.
Written by Jens Mueller, Senior Research Computing Specialist, Research Computing Support, Miami University.
Photos of Scott Hartley and DJ Rao by Jeff Sabo, Miami University Photo Services.
Miami has recently expanded its license for the Matlab product suite. Whereas just a few toolboxes were available under the previous license, Miami students, faculty, and staff now have access to about 80 different toolboxes. Application areas cover robotics, control systems, quantitative finance, computer vision, data analytics, deep learning and signal processing, and more. Along with the new license comes access to the full Academic Online Training Suite.