Technology Innovations Already Make the Aurora Exascale Supercomputer a National Asset

A massive, on-going, behind-the-scenes effort has already made the Aurora exascale supercomputer that will be built at Argonne National Laboratory a national asset. This effort includes over 80 software developers and 25 application teams, plus several science teams from universities and national laboratories participating in the Argonne Leadership Computing Facility’s Aurora Early Science Program.

In this blog, I’ll share a behind the scenes look at the technologies and the use cases that are behind the development of Aurora, a system that will give scientists an unprecedented set of capabilities that will enable this Intel-based system to be productive on the first day it is turned on.

2020 demonstrated how devastating health challenges such as the COVID-19 and climate change can be to families, society, and ecosystems around the world. Designed for exascale sized problems, Aurora has the computational capability to deliver urgently needed science to address big challenges and even assist decision makers in a timely fashion when managing time critical situations. Aurora will also open new vistas for established research efforts such as those in cancer and astrophysics to help us better understand the causes and treatments for the long-endured human nemesis of cancer, and to provide deeper insight into our universe.

Quick Facts about the Aurora Exascale system

Aurora is designed to augment human understanding with machine learning through the convergence of HPC, AI, and High Performance Data Analytics (HPDA). This system will showcase multiple new technologies at scale that will benefit future exascale efforts as well as demonstrate the efficacy of these technologies for enterprise and cloud HPC/AI/HPDA workloads.

The breadth of the research from simulation to AI augmented identification of cancer treatments is, in a word, breathtaking. The importance of the work is best understood through the words of the research scientists performing the work. Some short videos of representative research can be seen via the following links:

A key part of Aurora is the Intel® Optane™ persistent memory (Intel PMem) subsystem, which the CPU can access directly across the memory bus due to the innovative DIMM form factor packaging of the Intel PMem devices. This significantly reduces latency compared to a PCIe device and delivers new high-water marks in the amount of high-speed memory capacity and storage performance that scientists can use to explore new frontiers in big data, scientific computing, and AI applications. When used as a storage device, systems software such as DAOS (Distributed Asynchronous Object Storage) give applications direct access to Intel PMem storage objects from user space. The resulting reduction in overhead is already setting new performance records giving applications access to data with tens of gigabytes per second of memory bandwidth per node and with latencies measured in nanoseconds. Aurora’s performance characteristics will make possible previously intractable big data and data intensive computer-based scientific experiments tractable.

This revolutionary Intel memory subsystem has given our many extraordinary teams of Intel engineers the ability to create a balanced memory and storage architecture that will be unique to Aurora and its ability to run big data AI and huge unstructured data analytics jobs as well as deliver groundbreaking large sparse and dense matrix operation capabilities. Advances in software defined visualization (SDVis) with the Intel oneAPI rendering toolkit now give scientists the ability to photorealistically render data on the Aurora computational nodes - and keep the data in Intel Optane memory – to trigger on complex events and then interactively review the causes of the events.  With Intel Optane PMem devices (which can be used both as main memory or as a storage device), Intel is uniquely positioned to efficiently deliver the memory capacity and bandwidth to exploit the massive exascale compute capability and parallelism of the Intel many-core processors plus forthcoming Intel Xe GPUs.

Cross fertilization is an important benefit of the new Aurora-based Intel technologies. As I write this, people are experimenting with Intel Optane technology and the latest in Intel computational hardware. Even before becoming operational, the technology innovation of the Aurora exascale system makes it a strategic asset that is already having an impact in tackling challenges in scientific discovery, strengthening national security, and improving industrial competitiveness. All this exascale performance will be based on Intel’s future Sapphire Rapids CPU, discrete Xe GPUs (using Foveros and EMIB packaging technology), and Intel Optane memory-- all linked together with the oneAPI programming model. The research opportunities created by Aurora, are breathtaking.

Intel is inspired to create world changing new technology that enables global progress and enriches lives. As we move into the future, we are excited and committed to partner with Argonne National Laboratory in silicon development, future architectures for HPC and AI, and software enablement blazing uncharted discoveries fueled by Intel technologies designed for datacentric workloads.

 

For more information:

Intel technologies may require enabled hardware, software or service activation.

© Intel Corporation.  Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries.  Other names and brands may be claimed as the property of others.  ​

Published on Categories Artificial Intelligence, Big Data, High Performance Computing
Trish Damkroger

About Trish Damkroger

Patricia (Trish) A. Damkroger is vice president and general manager of the High Performance Computing organization in the Data Platforms Group at Intel Corporation. She leads Intel’s global technical and high-performance computing (HPC) business and is responsible for developing and executing strategy, building customer relationships and defining a leading product portfolio for technical computing workloads, including emerging areas such as high-performance data analytics, HPC in the cloud and artificial intelligence. An expert in the HPC field, Damkroger has more than 27 years of technical and managerial expertise both in the private and public sectors. Prior to joining Intel in 2016, she was the associate director of computation at the U.S. Department of Energy’s Lawrence Livermore National Laboratory where she led a 1,000-member group comprised of world-leading supercomputing and scientific experts. Since 2006, Damkroger has been a leader of the annual Supercomputing Conference (SC) series, the premier international meeting for high performance computing. She served as general chair of the SC’s international conference in 2014 and has held many other committee positions within industry organizations. Damkroger holds a bachelor’s degree in electrical engineering from California Polytechnic State University, San Luis Obispo, and a master’s degree in electrical engineering from Stanford University. She was recognized on HPC Wire’s “People to Watch” list in 2014 and 2018.