EE HPC WG 2022 Workshop

Program

Tuesday, May 17th 7:00AM - 10:00AM Pacific Time

David Patterson presents A Decade of Machine Learning Accelerators: Lessons Learned and Carbon Footprint, Video Recording

John Shalf presents Heterogeneous Integration Roadmap for the Future of Energy Efficient HPC, Video Recording

Sustainability panel moderated by Jim Rogers with panelists Steve Hammond, Johnny Dicus and Mark Monroe, Video Recording

"Liquid Cooling" panel moderated by Dave Martinez with panelists David Masuda, David Grant, Sammy Zimmerman and Steve Harrington, Video Recording

Wednesday, May 18th 8:00AM - 10:00AM Pacific Time

"Discussion Breakout Sessions (recording of recap from discussion leads)" Liquid Cooling led by Chris DePrater, Sustainability led by Steve Bruno, Energy Efficient HPC Software (recording) led by Matthias Maiterth and Operational Data Analytics (recording) led by Michael Ott

Natalie Bates presents Update on EE HPC WG Teams, Video Recording

A Decade of Machine Learning Accelerators: Lessons Learned and Carbon Footprint, by David Patterson

Abstract: The success of deep neural networks (DNNs) from Machine Learning (ML) has inspired domain specific architectures (DSAs) for them. We start the talk with lessons Google learned from designing and deploying many generations of ML DSAs since 2013, such as DNNs grow rapidly; the bottleneck is memory, not floating-point units; and semiconductor technology advances unequally. The rapid growth of DNNs rightfully raised concerns about their carbon footprint. The second part of the talk identifies the “4Ms” (Model, Machine, Mechanization, Map) that, if optimized, can reduce ML training energy by up to 100x and carbon emissions up to 1000x. By improving the 4Ms, ML held steady at <15% of Google’s total energy use despite it consuming ~75% of its floating point operations. Climate change is one of our most important problems, so ML papers should include emissions explicitly to foster competition on more than just model quality. External estimates have been off 100x–100,000x, so publishing emissions also ensures accurate accounting, which helps pinpoint the biggest challenges. With continuing focus on the 4Ms, we can realize the amazing potential of ML to positively impact many fields in a sustainable way.

Heterogeneous Integration Roadmap for the Future of Energy Efficient HPC, by John Shalf

Abstract: Advanced packaging architectures are today widely acknowledged as being increasingly important to drive performance and cost improvements of microelectronics systems. As a result, several innovative packaging architectures have been announced in recent years. On-package integration provides a compact, power efficient platform for Heterogeneous Integration of diverse IP that support faster time to market and cost/yield benefits. I will talk about the Heterogeneous Integration Roadmap activity, which promises another decade of energy efficiency and performance improvements.

Speakers and Organizing Committee

David Patterson received BA, MS, and PhD degrees from UCLA. He is a UC Berkeley Pardee professor emeritus, a Google distinguished engineer, the RIOS Laboratory Director, and the RISC-V International Vice-Chair. His most influential Berkeley projects likely were RISC and RAID. He received service awards for his roles as ACM President, Berkeley CS Division Chair, and CRA Chair and awards for his teaching. The most prominent of his seven books is Computer Architecture: A Quantitative Approach. He and his co-author John Hennessy shared the 2017 ACM A.M Turing Award, the 2021 BBVA Foundation Frontiers of Knowledge Award, and the 2022 NAE Charles Stark Draper Prize for Engineering. The Turing Award is often referred to as the “Nobel Prize of Computing” and the Draper Prize is considered a “Nobel Prize of Engineering.”

John Shalf is Department Head for Computer Science Lawrence Berkeley National Laboratory, and recently was deputy director of Hardware Technology for the DOE Exascale Computing Project. Shalf is a coauthor of over 80 publications in the field of parallel computing software and HPC technology, including three best papers and the widely cited report “The Landscape of Parallel Computing Research: A View from Berkeley” (with David Patterson and others). He also coauthored the 2008 “ExaScale Software Study: Software Challenges in Extreme Scale Systems,” which set the Defense Advanced Research Project Agency’s (DARPA’s) information technology research investment strategy. Prior to coming to Berkeley Laboratory, John worked at the National Center for Supercomputing Applications and the Max Planck Institute for Gravitation Physics/Albert Einstein Institute (AEI) where he was was co-creator of the Cactus Computational Toolkit.

Jim Rogers is the Computing and Facilities Director for the National Center for Computational Sciences (NCCS) at the Oak Ridge National Laboratory (ORNL). Mr. Rogers has thirty years of experience in high-performance computing and has provided strategic planning, technology insertion, and integration support for multiple computing centers, including ORNL's Leadership Computing Facility, the U.S. Air Force, NOAA's National Climate Computing Research Center (NCRC), U.S. Army Corps of Engineers Engineer Research and Development Center (ERDC), the Aeronautical Systems Center, the National Aeronautics and Space Administration (NASA) Goddard Space Flight Center, NASA Ames Research Center, the Defense Intelligence Agency, and the Alabama Supercomputer Center. He has primary responsibility for the strategy, acquisition, delivery, integration, and transition to production for high performance computing, storage, networking, and analysis systems as well as the physical facilities that house these systems. He manages recurring operational activities for both the NCCS systems and the supporting facility/infrastructure . These activities extend across multiple Federal customers.

Mark Monroe is a Principal Engineer in Microsoft’s Datacenter Advanced Development group working on future systems for energy, water conservation, and automation. He is a Six Sigma Master Black Belt, with a focus on change acceleration in corporate settings. Mark led industry consortium including Infrastructure Masons and The Green Grid, and served on the board of directors for the Center for ReSource Conservation in Boulder, CO. He has a Master of Science in Mechanical Engineering from Virginia Tech.

Steve Hammond is a Senior Research Advisor in the Mechanical and Thermal Engineering Sciences Directorate at DOE’s National Renewable Energy Laboratory. He leads efforts to address industry decarbonization and challenges in optimization of integrated complex systems, including smart urban systems, low-carbon integrated mobility, and catalyzing a clean energy transition. Steve was the founding director of the NREL’s Computational Science Center, where he was responsible for leading laboratory-wide efforts in computational science, high performance computing and energy efficient data centers.

John Dicus has provided over 37 years of federal service to this great country. He spent twenty years in the United States Air Force. John currently serves as Departmental Program Manager for IT Sustainability and Data Center Optimization and brings his leadership to the division providing the agency with oversight of the Data Center Optimization Initiative (DCOI). John has served as the Director of Customer Support Division for the DOE OCIO as well as the Director of Enterprise Service Center West for the DOE OCIO. John also served as Operations lead for NNSA Service Center in Albuquerque, NM. John has extensive management experience from the US Air Force in Network Operations Centers (NOC). Also while in the Air Force John worked with the Air Force Battle lab developing and implementing leading edge technologies. John’s information technology background is broad to include areas of systems administration, Systems engineering, facilities management, and operational cyber security. He currently received certification as a Data Center Energy Practitioner. John’s formal education includes a BS in Professional Aeronautics from the Embry-Riddle Aeronautical University. Two AS degrees in Computer Science and Armament Systems Technology from the Community College of the Air Force.

Dave Martinez is an Engineering Program/Project Lead and has worked in the Sandia National Laboratories Corporate Computing Facilities (CCF) for 35+ years. David is the subject matter expert for SNL’s data center operations and design due to his in-depth understanding and experience with HVAC, controls, and mechanical and electrical systems. David is frequently consulted by internal and external agencies for design review and his innovative approach to data center management and energy efficient operations and designs. SNL has received numerous energy efficiency awards as a result of these efforts. During his tenure, David has seen the data center operations move from about 20,000 sq. ft. to over 100,000 sq. ft. comprised of 4 unique data center environments which includes a Leed gold HPC data center.

Steve Harrington spun Chilldyne out of Flometrics in 2011 to increase the efficiency of data center cooling. He started Flometrics in 1995 while getting a PhD in Aerospace engineering with a focus on fluid dynamics. At Flometrics he helped design Laser cooling systems for ASML, medical ventilators for Respironics, UAV cooling systems for Northrop Grumman and Rocket fuel pumps for DARPA. Steve also taught Aerospace Engineering at UCSD, where his students designed built and flew liquid fueled rockets. He also has owned ski boats since 1982, rebuilding engines and outdrives, so he knows about long term effects of corrosion.

David Grant graduated from the University of Tennessee in 2003 with a B.S. in Mechanical Engineering. He has been at the Oak Ridge National Laboratory (ORNL) since 2009 and is a senior technical staff member. While at ORNL he has been involved with the design, construction, and operation of the mechanical systems supporting ORNL's 80,000SF+ of data centers which house the Summit and Frontier High Performance Computer Systems among others. He is currently a co-chair of the Energy Efficient HPC Working Group Infrastructure sub-team and is a corresponding member of the ASHRAE TC9.9. David is a registered Professional Engineer with the State of Tennessee and is a Certified Energy Manager (CEM - from the Association of Energy Engineers (AEE)) and a Data Center Energy Practitioner – Specialist (DCEP - from the Department of Energy (DOE)).

David Masuda graduated from Santa Clara University with a degree in Mechanical Engineering. He has spent the last 17 years working in Industrial water treatment with a focus on microelectronics and data centers. His hometown is Dublin, California where he and his young family enjoy fishing, hiking, surfing and traveling.

Matthias Maiterth is academic research staff at the Technische Universität München (TUM). Matthias obtained his Doctoral Degree from the Ludwig-Maximilians Universität München in September 2021 for his work on ``A Reference Model for Integrated Energy and Power Management of HPC Systems''. For his Doctoral Studies, he joined Prof. Kranzlmüller's research group at the LMU Munich in 2015. His studies include research exchanges with Lawrence Livermore National Laboratory during both his Master's and Doctoral program, as well as a three-year funded position at Intel for his Doctoral research in collaboration with Leibniz Supercomputing Centre (LRZ) in the area of energy- and power-aware runtimes systems in High Performance Computing. Matthias Maiterth joined the CAPS group of Prof. Martin Schulz in August 2021.

Michael Ott is a senior researcher in the “High-Performance Systems” division at Leibniz Supercomputing Centre (LRZ) and leads the research activities on energy efficiency. He is the technical lead of the Operational Data Analytics team of the Energy Efficient HPC Working Group (EE HPC WG) and leads the Energy Efficiency Working Group of the ETP4HPC. His research focuses on energy-efficient HPC operations, scalable monitoring, and Operational Data Analytics, but he still keeps up an interest in bioinformatics, computer architecture, and parallel programming.

Chris DePrater is a system engineer at Lawrence Livermore National Laboratory (LLNL) with over a decade of experience in High Performance Computing (HPC). Specializing in building controls, mechanical systems and is an active member Energy Efficient HPC working Group (EEHPCWG). Chris has a certification in Industrial Maintenance Air Conditioning and a bachelor’s degree From DeVry University in Electronic Engineering. He has been involved in the planning and siting of multiple top 500 HPC systems, Sequoia, Sierra and currently working towards Exascale at LLNL. Chris values teamwork and collaboration with peers in and out of the workplace. Outside of work Chris is a devoted father of 7 and enjoys the outdoors.

ORGANIZING COMMITTEE

Siddhartha Jana, Intel

Torsten Wilde, Hewlett Packard Enterprise

Anna Maria Bailey, Lawrence Livermore National Laboratory

Natalie Bates, Energy Efficient HPC Working Group

Chris DePrater, Lawrence Livermore National Laboratory

David Grant, Oak Ridge National Laboratory

Jason Hick, Los Alamos National Laboratory

Brandon Hong, Lawrence Livermore National Laboratory

Herbert Huber, Leibniz Supercomputing Center

Thomas Ilsche, Technical University Dresden

James Laros, Sandia National Laboratory

Steven Martin, Hewlett Packard Enterprise

Dave Martinez, Sandia National Laboratory

Jim Rogers, Oak Ridge National Laboratory

Dale Sartor, Lawrence Berkeley National Laboratory (retired)

Fumiyoshi Shoji, RIKEN

David Sickinger, National Renewable Energy Laboratory

Gert Svensson, KTH, Royal Institute of Technology

Michael Ott, Leibniz Supercomputing Centre

Gregg Barrett, Cirrus AI

Submission

This workshop features a keynote, interactive discussion groups as well as informational invited speakers and panels. There are no submissions, per se, however the EE HPC WG Organizing Committee is very interested in your feedback and suggestions. Please contact natalie@eehpcwg.org.

Speakers

Dave Patterson, Google and UC Berkeley

John Shalf, LBNL

Jim Rogers, ORNL

Mark Monroe, Microsoft

Steve Hammond, NREL

Johnny Dicus, US DOE

David Martinez, Sandia NL

David Grant, ORNL

Steve Harrington, Chilldyne

David Masuda, Chemtreat