Collaborative Supercomputing

By Michael Feldman

September 11, 2008

The big news in the science community this week was the kickoff of CERN’s Large Hadron Collider (LHC), The $10 billion atom smasher that sent its first proton beams through the device’s 17-mile underground tunnel in Switzerland and France. These initial tests were the culmination of 15 years of planning and development that brought together 80 countries and thousands of individual researchers around the world. While it remains to be seen what scientific discoveries will eventually result from the LHC experiments, there is no doubt it represents the biggest and most ambitious global science project today.

Today, though, I’m going to talk about another set of science community partnerships, although these have received much less attention from the press. For the past seven years, the U.S Department of Energy’s (DOE) Office of Science has opened the doors to its terascale supercomputers and changed the way many U.S. scientists are doing cutting-edge research. Through the SciDAC and INCITE programs, the Office of Science has expanded the high-end computing capabilities of the agency, while spreading supercomputing talent and hardware resources across the broader research community.

In most cases, these collaborations were confined to U.S.-based science, but in others cases, the DOE partnered with researchers from around the world. In fact, the DOE (along with the NSF) invested $531 million in the aforementioned LHC project and helped design and build the ATLAS and CMS detectors through two of its labs — Brookhaven in New York and Fermilab in Illinois.

Since U.S. government agencies compete for budget dollars and attention, the natural reaction is for each agency to guard its resources. So in many ways, the opening of the DOE’s Office of Science was an unlikely path. Perhaps even more unlikely is the individual who led the charge: Dr. Raymond Orbach, the director of the Office of Science. In 2002 Orbach was appointed by the Bush administration — a group not exactly known for its collaborative style of governance, much less its love for open science, or, in some cases, science at all. But Orbach proved to be an true leader in promoting partnerships with other agencies, universities and even industrial organizations.

The INCITE program, in particular, changed the nature of computing at the DOE. Up until 2002, agency computers were primarily used for DOE grantees. At that point, Orbach devised the INCITE program, which opened up DOE supercomputing resources to the science community. The program was designed so that supercomputing cycles were allocated on a competitive basis, in which only the most capable organizations and the most interesting problems were given time on the machines. In a nutshell, the idea was to make the best hardware available for the best science. “It sounds completely reasonable now, but I can tell you back in 2002, there was lot of speculation and complaints that I was opening up our computers to the world,” admits Orbach.

In each succeeding year the program expanded its allocations. In 2008, 265 million CPU hours on DOE machines were awarded to 55 projects, eight of which are from industry, 17 from universities, and 20 from DOE labs as well as other public, private and international researchers.

As it turns out, INCITE will also provide the structure for computer allocations announced on Monday for a new partnership with NOAA. In this case, the Office of Science will make available more than 10 million hours of computing time for NOAA to develop and refine advanced climate change models. The work will be performed on the latest computing hardware at three DOE labs: Argonne, Oak Ridge, and NERSC at Lawrence Berkeley.

Although the DOE has worked with the climate community before, it’s mostly been done via lower-level collaborations between PIs across agencies. As Pete Beckman, Argonne’s Interim Director, puts it: “This really says we want to move together in a strategic way. And that’s very important.” He sees the new collaboration as a way to move the national climate and weather modeling work forward under a more unified structure. At Argonne, they’ve already begun porting some of the NOAA codes to run on their 557-teraflop Blue Gene/P system. Under this new framework, Beckman believes over the next couple of years we should be able to “dramatically improve our capabilities for weather and climate prediction.”

The collaboration between the DOE and NOAA has been formalized in a memorandum of understanding (MOU), but is being done under the general framework of the Climate Change Science Program (CCSP), which was instituted in July 2003. The program brought together not just NOAA and the DOE, but also NCAR and ten other federal agencies. The rationale was to bring some cohesion to the climate codes being developed across the United States. “At the time the CCSP was promulgated, the United States was behind in high-end computation,” says Orbach. “The Japanese Earth Simulator was the fastest machine in the world and we didn’t have any open science capability to match it.”

In 2004, the U.S. took back the top spot in supercomputing with BlueGene/L and has maintained the lead ever since. But not all U.S agencies were endowed with leading-edge supercomputers or the software talent they attracted. The NOAA, which is administered by the Department of Commerce, has much less HPC capability than federal agencies like the Department of Defense and the DOE. Currently, the most powerful system owned by NOAA is a relatively modest 25-teraflop IBM Power6-based system. “In my view, this MOU is a recognition of where each of the agencies is at this point in time, and frankly a rationalization of their capabilities and talents,” explained Orbach.

In addition to the retaking of supercomputing leadership in 2004, the climate modeling community also expanded. Through the Atmospheric Radiation Measurement (ARM) and the SciDAC program (a program begun in 2001 that brought together top researchers in a variety of scientific disciplines), the Office of Science has developed deep expertise in both climate measurement and global climate change modeling. According the Orbach, the accumulation of this expertise over the last five years is at least as important as the new hardware in moving the climate models forward.

Under the new partnership, the NOAA codes will become open to the community, and Orbach is hoping that the software will be optimized with the help of SciDAC researchers. NOAA currently uses one of its home-grown Geophysical Fluid Dynamics Laboratory (GFDL) codes to predict hurricanes, but it is currently limited to a grid model based on 9 km granularity. Orbach says to get really accurate models, you need to get down to the 1 km level. By optimizing the software, Orbach thinks you can pick up a couple orders of magnitude in “effective speed,” and notes that SciDAC has made similar improvements in other codes they’ve worked on.

NOAA’s GFDL climate codes and the DOE-NCAR Community Climate System Model (CCSM) codes are the two national major climate models developed in the U.S. The CCSM model is already being run extensively on DOE supercomputers, while porting of the GFDL code is imminent. Whether the codes become integrated at some point or continue to diverge remains a question, although Orbach has his own take on this.

“Ultimately, I would like to see the United States have a single code,” he says. “That’s what the Europeans have agreed on and they have many more partners than we have. As a consequence, they’ve been able to develop a common code for the whole European community and have made really wonderful advancements. This multiplicity of codes — I don’t know how it’s going to shake out.”

Looking further out, Orbach would like to see the climate change models incorporate human factors. Today the climate codes only take into account the physical system — the oceans, the atmosphere, land masses, etc. But human behavior can be modeled as well, and since people will necessarily change their behavior in response to climate policy decisions — for example, energy pricing, new energy sources, and conservation measures — that feedback must be part of the climate model to produce an accurate prediction. More importantly, the policy makers themselves would need access to those models so they could run different scenarios for policies they are considering.

Integrated models like this are already being talked about in anticipation of multi-petaflop and eventually exaflop DOE machines. The agency has already donated a million CPU hours to the National Endowment for the Humanities (NEH) to begin to generate interest in this type of application. Also, workshops at Berkeley have been set up to teach social scientists how to make use of these leading edge supercomputers.

“We’re not there yet,” Orbach told me. “I don’t know how fast our computers are going to have to be or how good our codes are going to have to be, but you can see where we’re going.”

Orbach probably won’t be around to usher in these next-generation applications though. As a political appointee, his tenure at the DOE ends in four months, when Bush leaves office. He assumes that whoever prevails in the presidential election will want their own person to head the Office of Science. “I disappear at noon on January 20th,” notes Orbach.

Theoretically, his INCITE program could be ditched or scaled back by new leadership, but that’s highly unlikely. The program is already too popular in the science and technology community. What’s more likely is that the next director will build on the foundations that Orbach has built over this 7-year tenure. And with public awareness of climate change and energy policy at an all-time high, the DOE may well be the most important agency of the U.S. government in the next administration.

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