Driving Scientific Discovery and Strengthening Competitiveness with Cutting-edge Computational Tools

The breadth and depth of research supported by Advanced Research Computing (ARC) at the University of Michigan is nothing short of monumental. ARC provides more than 3,000 users with the high performance computing (HPC) resources and support they need to engage in computational and data-intensive scientific investigations across a wide range of disciplines. For example, ARC is enabling new interdisciplinary research projects, helping scientists conducting artificial intelligence and machine learning research, by providing expanded computing infrastructure, consulting expertise and support.

Whether it’s providing access to HPC systems, consulting on research methods or assisting with funding submissions, ARC is there to help U-M researchers drive scientific discovery. In the process, ARC helps strengthen the competitiveness of the academic and research programs at U-M, an institution with more than 40,000 undergraduate students and over 100 graduate programs ranked in the Top 10.

To carry out its expansive mission, ARC regularly adds new HPC hardware and software resources that keep leading-edge computational tools in the hands of the university’s research community. That’s the case with U-M’s new campus-wide computing cluster, known as Great Lakes.

The Great Lakes cluster, which went into production in July, arms the university’s research community with the power of approximately 15,000 compute cores. The next-generation system, built on Dell EMC-enabled HPC infrastructure, is designed to provide a balanced combination of computing power, I/O performance, storage capability and accelerators.

The building blocks of the cluster include:

• Dell EMC PowerEdge™ C6420 compute nodes, PowerEdge R640 high-memory nodes and PowerEdge R740 accelerator nodes
• Mellanox HDR 200 Gb/s InfiniBand ConnectX-6 adapters, Quantum switches and LinkX cables, and InfiniBand gateway platforms
• DDN GRIDScaler® 14KX® storage system and 100 TB of usable IME® (Infinite Memory Engine) memory

In a notable advance, Great Lakes is the industry’s first production system to benefit from Mellanox HDR 200 Gb/s InfiniBand networking, which enables faster data transfer speeds and increased application performance.¹ And for even more performance, the cluster incorporates accelerators designed to speed up machine learning, deep learning, graphics and various other HPC workloads.

Great Lakes is a highly versatile system. It’s designed to serve the diverse needs of researchers across the university, as opposed to supercomputers built to handle a narrower range of workloads. The Great Lakes cluster is available to all researchers on the U-M campus for simulation, modeling, machine learning and natural language processing, data science, genomics, bioscience and more.

The use cases for the system are all over the map. Health-related activities include research in sports medicine, precision medicine, genetics and public health. Other research is focused on 5G networking, instrumented sensors in “smart cities,” preventative maintenance on automotive manufacturing lines and their robotic components, civil engineering and urban infrastructure, chemistry, music, driverless cars, and much more.

In designing the Great Lakes cluster, the ARC Technology Services team worked closely with Dell EMC and its technology partners. One of the goals of this collaborative effort was to build a balanced system that would give different researchers the performance they need without over-investing in costly technologies. In an example of this balanced approach, the Great Lakes architecture puts 100 TB of high-performance NVMe flash memory in front of 2 PB of spinning disk to transparently accelerate workloads that are I/O bound, and to do so without having to build the entire system in costly flash memory.

Ultimately, Great Lakes is a unique system that is helping the growing U-M research community accelerate their scientific investigations while enhancing teaching and learning across the campus.

To learn more

• For the full story, read the Dell EMC case study “‘Great Lakes’ Cluster: The University of Michigan enhances its research program with a next-generation supercomputer.”
• To learn more about how U-M is providing crucial ingredients for success in AI and machine learning, read U-M fosters thriving artificial intelligence and machine learning research.
• To explore Dell EMC PowerEdge servers with Intel Xeon processors in HPC applications, visit Dell EMC solutions for HPC and AI.
• For perspectives on tapping the value of data with deep learning and artificial intelligence systems, explore Dell Technologies AI Solutions and Dell EMC Ready Solutions for AI.

1. University of Michigan news release, “U-M selects Dell EMC, Mellanox and DDN to Supply New “Great Lakes” Computing Cluster,” October 16, 2018.