Fixating on Exascale Performance Only Is a Bad Idea

By John Russell

June 15, 2015

Few in the computing industry dispute that achieving ‘exascale’ performance is worthy on its own merits. But fixating only on the magic milestone (10^18 FLOPS) misses an important point – HPC as currently practiced, even at the exascale level, won’t solve the grand challenges in science and society contends Dr. Sadasivan Shankar, Harvard University visiting lecturer and long-time prominent Intel materials scientist.

“I would argue that exascale is more than just exascale. It’s a new way of thinking. I would even argue that [when] petascale came along a lot of us who were using large scale computing didn’t see the benefits. If we are planning to solve some of the large scale grand problems in science and society today, the premise I am putting forth is we should do it a little differently,” Dr. Shankar told the audience at the recent HPC User Forum.

The new evolving paradigm, at least in materials science, is In Silico Inverse Design, proposed Dr. Shankar in his presentation – Exascale for Grand Challenge Problems in Sciences and Engineering. He presented his case for inverse design by examining emerging disruptive technologies in both science and society, identifying commonality in the challenges they represent (e.g. scale and combinatorial complexity), and presented a small sample of related computational challenges.

Many in the computer industry may know Dr. Shankar, who is the first Margaret and Will Hearst Visiting Lecturer in Computational Science and Engineering at Harvard School of Engineering and Applied Sciences. He was formerly an Intel scientist working in semiconductor materials, design and manufacturing.

During his tenure in the semiconductor industry Dr. Shankar worked on initiatives using computational modeling to optimize semiconductor processing and equipment for several technology generations, advanced process control using physics-based models, thermo-mechanical reliability of microprocessors, thermal modeling of 3D die stacking, and using thermodynamic principles to estimate energy efficiency of ideal computing architectures. He is a co-inventor in several patent filings covering areas in chemical reactor designs, semiconductor processes, bulk and nano materials, device structures, and algorithms (fuller bio).

The core of his argument is that in silico inverse design represents an emerging paradigm characterized by distinct computational requirements that are different and challenging as compared to direct design. Not a new concept, inverse design has gained traction in recent years. The Department of Energy’s Center for Inverse Design offers the following description:

“…To address a crucial scientific grand challenge…rather than using the conventional direct approach (Given the structure, find the electronic properties), we are using a “materials by inverse design” approach (Given the desired property, find the structure).

“The target properties of interest include general semiconductor optical and electrical properties; the desired materials functionalities include electron- and hole-conductive transparent conductors, solar absorbers, and nanostructures for energy sustainability. Our predictions of materials are examined iteratively by various synthetic approaches, including high-throughput parallel materials science.”

Dr. Shankar’s succinct definition is “the ability to use predictive capabilities, to design a material, which when synthesized in the manufacturing line will exhibit targeted properties.” He is quick to add, because of the approach “It’s not always guaranteed there is, in fact, a solution.”

Screen Shot 2015-06-14 at 6.02.55 PMA nuanced theory perhaps best heard directly from Dr. Shankar (link to a video of his presentation is at the end of the article); it is nonetheless possible to outline some of the central ideas he presented. Number one, science and technology have advanced to the point, he contends, that the direct design approach no longer works as well. Number two, the disruptive challenges facing both science and society share important technology commonalities best addressed by inverse design.

He cited work (McKinsey) that identified the top technology-based disruptive forces in the worldwide economy and business as: mobile internet, automation, internet of things, advanced robotics, autonomous and non autonomous vehicles, next-gen genomics, energy storage, 3d printing, advanced materials, advanced oil and gas exploration, and renewable energy.

“Traditionally disruptions are caused by emergence of a different kind of material, different kind of process, or a different way of doing something. Those cited here are expected to completely upstage the worldwide economy by 2025,” he said.

DOE offers a list of five expected scientific disruptors: control of material processes at the level of atoms and electrons; design and perfect atom- and energy- efficient synthesis of revolutionary new forms of matter with tailored properties; emergence of remarkable properties of matter from complex correlations of the atomic or electronic constituents and optimal control of these properties; characterization and control matter of non-equilibrium systems; leverage knowledge of energy and information interactions from living beings at the nanoscale to design systems with high energy efficiencies.

“If you look at these challenges (science and business) side by side and look for commonalities you could distill them to five items: 1) design, 2) materials and devices, 3) sensing, 4) automation and 5) miniaturization, fabrication and testing,” he said, all of which are well suited to inverse design.

Leaving science aside for a moment, Dr. Shankar noted the political landscape is shifting in support of funding supercomputing generally and exascale in specific (See Obama’s 2016 Budget Boosts R&D, Exascale Funding, HPCwire. The Partnership for Advanced Computing in Europe (PRACE) is also chasing exascale. China, of course, has the fastest supercomputer on the TOP500 list.)

Amid the rush to reach exascale Dr. Shankar believes the practicality of turning a quintillion FLOPS into real-world applications is perhaps not getting enough attention. The race to exascale (and resulting technologies and architectures) should be well informed by real-world applications and requirements of inverse design methodology – nowhere more so than in material science which has such a direct effect on economies and progress.

In semiconductor materials synthesis generally there are at least six challenges said Shankar: “One is the ability to measure; the second is the ability to synthesize; the third is the intrinsic disparity in scales; the fourth is being able to compute as needed; the fifth is complexity; and sixth is the daunting combinatorics associated with trying to go from a small scale to a large scale.”

Screen Shot 2015-06-14 at 5.53.51 PMConsider the combinatorial challenge associated with semiconductor material synthesis.

“If you look at the number of elements you could use it’s about 60 to 70, basically anything that’s not toxic. If you look at how you can mix them up into alloys or compounds, there are about 100,000. Now if you take two of these materials and try to create a junction now the possibilities are in the billions,” said Dr. Shankar. This number quickly grows to trillion and beyond as the number of interfaces between structures grow.

“In my talk, I was trying to synthesize 4 complex premises [in addressing these challenges],” said Dr. Shankar:

  • Computational Design of Materials and Chemistry and why computing can play a key role
  • Sub-points for bio and systems biology also benefit
  • Eco-system is changing in favor of computing, both in terms of needs and political support
  • Disruptions in technology and fundamental sciences could be connected by computing

“Computing will be connecting the dots but not computing the way we do traditional compute,” he said. Here is a link to the full video of Dr. Shanker’s presentation.

 

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industy updates delivered to you every week!

NREL ‘Eagle’ Supercomputer to Advance Energy Tech R&D

August 14, 2018

The U.S. Department of Energy (DOE) National Renewable Energy Laboratory (NREL) has contracted with Hewlett Packard Enterprise (HPE) for a new 8-petaflops (peak) supercomputer that will be used to advance early-stage R&a Read more…

By Tiffany Trader

Training Time Slashed for Deep Learning

August 14, 2018

Fast.ai, an organization offering free courses on deep learning, claimed a new speed record for training a popular image database using Nvidia GPUs running on public cloud infrastructure. A pair of researchers trained Read more…

By George Leopold

CERN Project Sees Orders-of-Magnitude Speedup with AI Approach

August 14, 2018

An award-winning effort at CERN has demonstrated potential to significantly change how the physics based modeling and simulation communities view machine learning. The CERN team demonstrated that AI-based models have the Read more…

By Rob Farber

HPE Extreme Performance Solutions

Introducing the First Integrated System Management Software for HPC Clusters from HPE

How do you manage your complex, growing cluster environments? Answer that big challenge with the new HPC cluster management solution: HPE Performance Cluster Manager. Read more…

IBM Accelerated Insights

Super Problem Solving

You might think that tackling the world’s toughest problems is a job only for superheroes, but at special places such as the Oak Ridge National Laboratory, supercomputers are the real heroes. Read more…

Rigetti Eyes Scaling with 128-Qubit Architecture

August 10, 2018

Rigetti Computing plans to build a 128-qubit quantum computer based on an equivalent quantum processor that leverages emerging hybrid computing algorithms used to test programs and potential applications. Founded in 2 Read more…

By George Leopold

NREL ‘Eagle’ Supercomputer to Advance Energy Tech R&D

August 14, 2018

The U.S. Department of Energy (DOE) National Renewable Energy Laboratory (NREL) has contracted with Hewlett Packard Enterprise (HPE) for a new 8-petaflops (peak Read more…

By Tiffany Trader

CERN Project Sees Orders-of-Magnitude Speedup with AI Approach

August 14, 2018

An award-winning effort at CERN has demonstrated potential to significantly change how the physics based modeling and simulation communities view machine learni Read more…

By Rob Farber

Intel Announces Cooper Lake, Advances AI Strategy

August 9, 2018

Intel's chief datacenter exec Navin Shenoy kicked off the company's Data-Centric Innovation Summit Wednesday, the day-long program devoted to Intel's datacenter Read more…

By Tiffany Trader

SLATE Update: Making Math Libraries Exascale-ready

August 9, 2018

Practically-speaking, achieving exascale computing requires enabling HPC software to effectively use accelerators – mostly GPUs at present – and that remain Read more…

By John Russell

Summertime in Washington: Some Unexpected Advanced Computing News

August 8, 2018

Summertime in Washington DC is known for its heat and humidity. That is why most people get away to either the mountains or the seashore and things slow down. H Read more…

By Alex R. Larzelere

NSF Invests $15 Million in Quantum STAQ

August 7, 2018

Quantum computing development is in full ascent as global backers aim to transcend the limitations of classical computing by leveraging the magical-seeming prop Read more…

By Tiffany Trader

By the Numbers: Cray Would Like Exascale to Be the Icing on the Cake

August 1, 2018

On its earnings call held for investors yesterday, Cray gave an accounting for its latest quarterly financials, offered future guidance and provided an update o Read more…

By Tiffany Trader

Google is First Partner in NIH’s STRIDES Effort to Speed Discovery in the Cloud

July 31, 2018

The National Institutes of Health, with the help of Google, last week launched STRIDES - Science and Technology Research Infrastructure for Discovery, Experimen Read more…

By John Russell

Leading Solution Providers

SC17 Booth Video Tours Playlist

Altair @ SC17

Altair

AMD @ SC17

AMD

ASRock Rack @ SC17

ASRock Rack

CEJN @ SC17

CEJN

DDN Storage @ SC17

DDN Storage

Huawei @ SC17

Huawei

IBM @ SC17

IBM

IBM Power Systems @ SC17

IBM Power Systems

Intel @ SC17

Intel

Lenovo @ SC17

Lenovo

Mellanox Technologies @ SC17

Mellanox Technologies

Microsoft @ SC17

Microsoft

Penguin Computing @ SC17

Penguin Computing

Pure Storage @ SC17

Pure Storage

Supericro @ SC17

Supericro

Tyan @ SC17

Tyan

Univa @ SC17

Univa

  • arrow
  • Click Here for More Headlines
  • arrow
Do NOT follow this link or you will be banned from the site!
Share This