Making Sense of HPC in the Age of Democratization

By Tiffany Trader

March 8, 2016

These are exciting times for HPC. High-performance computing and its cousin high-productivity computing are expanding such that the previous definitions of HPC as a moving performance target or as the purview of modeling and simulation are breaking down. The democratization of HPC has spurred a lot of focus on the impact that HPC-hatched technologies are having on business computing, but HPC “proper” is also in the midst of a transformation.

To get some perspective on HPC’s shifting market dynamics and the confluence playing out at the intersection of HPC, big data and cloud, we reached out to Ed Turkel, an industry veteran and Dell’s HPC strategist.

The cornerstone of Dell’s solution model, according to Turkel, is “selling systems to democratize HPC.”

“There’s always going to be the customers at the high-end of HPC who are going to drive the technology, who are going to want the absolute most performance that they can get, and are always going to drive the marketplace in that direction,” Turkel states. “But what we are seeing at the other side of the market is a real broadening of the HPC market simply as more and more people use HPC technology to get their work done where there work isn’t traditionally something we associated with HPC.”

Dell’s solution approach to sales, which is tightly aligned with Intel’s Scalable System Framework, targets a class of user that values ease-of-use and ease-of-deployment over absolute maximum performance. Dell does the upfront work to determine what the design of the solution should be, what the software stack should be, and what hardware configuration should be in place. For many of these customers, sizing considerations take precedence over specific speeds and feeds.

“When we talk to the big government labs, they want to know how can we run one big job across thousands and thousands of cores, but that’s not the mainstream of HPC,” Turkel contends. “The mainstream of HPC is running much smaller jobs. You go to the automotive companies, even the tier one automotive companies, for example, that have some fairly large systems, and if you ask them how they use it, they tend not to run single jobs on huge numbers of systems — they have kind of a sweet spot that is the right balance of performance of the application, the cost of the software licenses — remember that a lot of the ISV codes charge by the thread. So there is a sweet spot. If you are not going to use the whole machine, running in a virtualized environment can be very efficient.”

These mid-range users are pulling in more enterprise-hardened technologies and tools to enable more agility, more ease of use, and greater flexibility. Turkel mentions the Tokyo Institute of Technology as a case in point. The Tsubame 2.0 HP machine could be dynamically provisioned to create a pool of virtual supercomputing nodes. The approach gave users the flexibility to run jobs across the system as with a traditional close to the metal environment, or to leverage the virtualized environment for a range of tasks. With virtual machines or today’s lighter-weight container technology, resources can be provisioned to support applications beyond the purview of traditional HPC.

The democratization of HPC has meant that Dell increasingly talks to a different class of user than previously. There are more medium-sized manufacturers, for example, where the volume of HPC work is not enough to keep their systems busy 24-7, so a shared workload approach enables them to do administrative and HPC workloads on one system. Turkel points out that while some of these customers may consider outsourcing to a public cloud provider, at the end of the day there’s enough HPC need to justify the expense of dedicated hardware if they can also use it for administrative tasks.

“More tier one or larger customers will tend to have distinct environments for their HPC work and for their business/administrative tasks and so on,” notes Turkel. “For mid-range customers, they may simply find it more economical to share and that gives them the advantage of having some bigger systems that they can use off-hours for running bigger HPC jobs but during the day take advantage of having all that gear and leverage for their administrative tasks — it’s an interesting model that I think for small and medium businesses that can be a very efficient way of using their systems — they are not the ones that are running these really big jobs that might be consuming their entire system that absolutely needs to be close to the metal because it needs 99.9 percent of the performance potential of the hardware. We dub this high-productivity computing because it’s a very efficient model. Just like virtualization, the goal is better utilization of the gear.”

Turkel cites as a recent example, an academic customer in Latin America. The mid-size university wanted to run all their administrative tasks including the library during operating hours, but then have the system’s full resources available for HPC the rest of the time. A mixed-workload approach made sense in this case because the university isn’t doing enough HPC to justify a dedicated environment, notes Turkel.

While the mid-range academic and industrial customer have some common needs around ease-of-use and ease-of-deployment, distinctions arise in terms of applications and workflows. “If you look at one of the big automotive companies, for example, they tend to run mostly the same set of applications (structural analysis, fluid dynamics, etc.) and they tend to run them a lot,” explains Turkel. “You go into a research environment and on any given day, the mix of applications they are running can be very different, depending upon what the research protocols going on at the university are. It will change from day to day, week to week.”

Because of the application set and the wide user base, the research community tends to require more dynamic and more agile computing environments than is typically true from the large industrial customers.

“For example, the Comet system at San Diego Supercomputing Center is actually set up somewhat as a cloud,” says Turkel. “Because the nature of the academic environment is that a researcher will want to run some set of jobs for the period of time that they are doing a particular research protocol and then give all that up and go on to do something else — so the ability to be able to get access to a part of the system for the purpose of running a particular set of applications and do that in a very dynamic way, makes much more sense in the research environment than is true necessarily in industrial accounts.

“Industrial accounts generally tend to be much more homogeneous in the things that they are doing on a day to day basis — typically running the same sorts of applications. This can change with projects spinning up and down, so they are also looking for agile environments, just not at the same level. It’s not like a professor who will want to run a set of simulations that might take 2-3 weeks; a project for a new vehicle design at an automotive company is something that would take months or even years.”

There’s also been an influx of applications using HPC technology for analytics, alone or in combination with modeling and simulation. “That’s occurring everywhere,” says Turkel, referring to the explosion in data analytics. “The percentage of time that we’re seeing a customer is purely doing traditional HPC is decreasing wildly.”

Life sciences exemplifies this trend, says Turkel. As we move into the age of personalized medicine, scientists are taking rich sets of molecular modeling data and genomics data and analyzing that in relation to clinical data to guide medical treatments and care.

Another example is financial services. Says Turkel: “They are still doing risk analysis, the typical Monte Carlo workloads, but they are also looking at how they can do analytics on all of the transactions that they’ve done with their customers over the years to look at buying patterns. It starts blending the traditional raw modeling/simulation/compute elements with using compute on vast bodies of data in an analytics sense for big data applications — that’s simply becoming more and more common as we look across the different vertical application spaces.”

(Re)defining HPC — It’s complicated

The more ubiquitous HPC becomes the more challenging it is to define. Dell has continued to evolve the in-house training material that it provides for sales and solution architects. “The slide that defines HPC has gotten much more complicated,” Turkel shares. “It used to be pretty simple, modeling and simulation — end of story. Now we talk about traditional HPC with modeling and simulation and we talk about data-centric HPC, which is like HPC but done on larger volumes of data — and we lump seismic analysis and genomics into that — and then high-performance data analytics starts getting into something else.”

With a field like personalized medicine, there is overlap, but then there are new applications, like fraud analysis, notes Turkel. This big analytics application is using HPC technology to prevent transaction fraud. Marketing analytics is another very new phenomenon. This is what allows Amazon to make suggestions to guide your next purchase and it’s why when you search for a gift for your cousin’s wedding, your Web surfing is suddenly dominated by ads for wedding products and services.

Interestingly, according to Turkel, even the use of traditional HPC is becoming more pervasive, driven by economics and the need for competitive advantage.

“When I started doing HPC many moons ago, you looked at the TOP500 list, for example, and you didn’t see too many industrial customers, and you didn’t see too many customers outside of so-called first-world nations, the US, Europe and so on. Now you look at it and it’s like the United Nations — it’s much more economical, so it’s being deployed everywhere. It’s being seen much more as a competitive weapon, as a way for countries, universities and businesses to invest in innovation.”

All of these trends — increased industry use, increased geographic representation and increased use of HPC for analytics – are reflected on the TOP500, says Turkel. “The technology is becoming more pervasive, the use of it is becoming more widely seen as necessary for businesses to be able to innovate.”

HPC’s role as an economic engine is increasingly recognized at the highest levels. This was the message that came out of the White House and the National Strategic Computing Initiative last summer, notes Turkel. “The government is really investing in HPC, not only for the DOD and DOE labs, but also just generally to make the technology more available. It has material impact on everything from healthcare to pick your products. And the analytic side of it is only causing it to grow that much more. As businesses seek to be able to do things like marketing analytics — how far away can you get from traditional HPC than marketing analytics, which is everywhere now.”

“I think what we’re seeing is a natural expansion of the market as it matures,” says Turkel. “As the technology matures, as it becomes more economical, as it becomes more approachable, you will see entities who a few years ago never would have thought of having done anything with HPC are now realizing that this is technology that they need to use to become competitive and innovate in the ways they want to innovate. The democratization and pervasiveness go hand-in-hand. The work being done to democratize HPC – to make it more affordable, more approachable, more easy to use, easy to buy, easy to deploy and so on – is pushing the pervasiveness.”

Subscribe to HPCwire's Weekly Update!

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

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, codenamed Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire