Keeping Big Data Cool at SDSC

June 29, 2016

June 29 — When most people think of a supercomputer center, they may think of one massive computer performing a single task. Inside the data center at the San Diego Supercomputer Center (SDSC) at the University of California San Diego, however, there are several large supercomputer systems, each performing multiple tasks simultaneously across a wide range of science domains that include genome sequencing to help pave the way to personalized medical treatment, coming up with new drug designs for conditions such as Parkinson’s and Alzheimer’s disease, or creating detailed fluid dynamics simulations for hypersonic aircraft.

Keeping SDSC’s main data center cool enough so that its Comet and Gordon supercomputers, among smaller clusters, don’t overheat is a complex yet mission-critical task, according to Todor Milkov, SDSC’s senior project engineer. A computing architecture such as the one found in Comet, SDSC’s newest supercomputer, requires one megawatt of power to operate the system. Using that much electricity generates a tremendous amount of heat, so SDSC, with the help of outside experts, developed three cooling system prototypes and conducted research to determine the most efficient system.

Each prototype system was designed using vendor-specific technology controlling five air handlers as a baseline to evaluate system performance. One of the prototypes used wireless temperature sensors that read the temperature of the hot and cold aisles every three minutes to increase battery life.

SDSC Datacenter AisleMany data centers use a standard hot aisle/cold aisle design. This design involves lining up server racks in alternating rows, with cold air intakes facing one way and hot air exhausts facing the other. The rows composed of rack fronts are called cold aisles. Typically, cold aisles face air conditioner output ducts. The rows that the heated exhausts pour into are called hot aisles. Typically, hot aisles face air conditioner return ducts.

Containment systems can help isolate hot aisles and cold aisles from each other and prevent hot and cold air from mixing. Such systems started out as using physical barriers that simply separated the hot and cold aisles with vinyl plastic sheeting or Plexiglas covers. Modern containment systems offer plenums and other commercial options that combine containment with variable fan drives (VFDs) to prevent cold air and hot air from mixing.

At SDSC, however, the entire area under the raised floor is used for the supply plenum, and the entire area above the ceiling is for the return plenum. Cold aisles use perforated floor tiles with specifically designed hole sizes to control the air flow volume from the space below the floor, while the hot aisles use ceiling grates that allow heated air to enter the space above the ceiling.

Controlling the air flow from all air handlers discharging into one common plenum presents a difficult problem, especially since these spaces also contain obstructions such as pipes and conduits. Moreover, not all of the compute clusters run at full capacity at any given time, and systems loads also change regularly as research projects start up or stop. These constantly changing factors cause the amount of heat dissipated from the supercomputer systems to fluctuate from minute to minute. The data center cooling system has to quickly adjust to accommodate these fluctuations in temperature.

“We learned a lot during the prototype and research phase of the cooling system design,” said Milkov. “We started by collecting a lot of data on how air flowed through the data center. We found that three minutes between temperature readings was too long an interval to keep the data center within the desired temperature ranges. Because of the longer interval, we used more electricity bringing the data center back to its temperature set points than we needed if we took temperature readings over shorter intervals and could make changes to the cooling system sooner.”

Realizing that a different approach was needed, Milkov put together a vendor evaluation process for an updated data center management system with the objective of reducing energy use while increasing the level of control capability available to the SDSC operations staff.

After extensive research, Milkov selected three companies for prototype installations. At the conclusion of a detailed evaluation, systems integration company Earth Base One (EBO) Corporation and a SNAP PAC-based control system were chosen for providing extensive control capabilities and energy savings.

Milkov and Michael Hyde, EBO’s president, approached the project with the same vision. “Rather than adapting an off-the-shelf data center management system to SDSC, we designed a tailor-built system for SDSC’s unique challenges,” said Hyde.

Opto 22, which develops and manufactures hardware and software products for applications in industrial automation, remote monitoring, and data acquisition, was chosen as the primary controls manufacturer. “The Opto 22 hardware and software not only won the competition for control and energy savings, but was also the least expensive vendor solution,” said Hyde. “The software’s excellent historical data collection and trending abilities allowed SDSC engineers to continue improving the system based on real data.”

“We appreciated the outstanding technical support SDSC received from Opto 22 during our design and prototype phase,” said Milkov. “When you’re trying to protect millions of dollars’ worth of research, you need a control system you can rely on.”

The full case study is available here.

About SDSC

As an Organized Research Unit of UC San Diego, SDSC is considered a leader in data-intensive computing and cyberinfrastructure, providing resources, services, and expertise to the national research community, including industry and academia. Cyberinfrastructure refers to an accessible, integrated network of computer-based resources and expertise, focused on accelerating scientific inquiry and discovery. SDSC supports hundreds of multidisciplinary programs spanning a wide variety of domains, from earth sciences and biology to astrophysics, bioinformatics, and health IT. SDSC’s Comet joins the Center’s data-intensive Gordon cluster, and are both part of the National Science Foundation’s XSEDE (eXtreme Science and Engineering Discovery Environment) program, the most advanced collection of integrated digital resources and services in the world.

About Opto 22

Opto 22 develops and manufactures hardware and software products for applications in industrial automation, remote monitoring, and data acquisition. Using standard, commercially available Internet, networking, and computer technologies, Opto 22’s input/output and control systems allow customers to monitor, control, and acquire data from all of the mechanical, electrical, and electronic assets that are key to their business operations. More information is at www.opto22.com.


Source: SDSC

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!

Microsoft, Quantinuum Use Hybrid Workflow to Simulate Catalyst

September 13, 2024

Microsoft and Quantinuum reported the ability to create 12 logical qubits on Quantinuum's H2 trapped ion system this week and also reported using two logical qubits on an H1 system to simulate an iron catalyst's low ener Read more…

Diversity Hiring Maximizes Everyone’s Success in STEM and Beyond

September 12, 2024

Despite overwhelming evidence, some companies remain surprised by this simple revelation: Diverse workforces and leadership teams are good for business. Companies that cultivate diverse hiring practices and maintain a di Read more…

GenAI: It’s Not the GPUs, It’s the Storage

September 12, 2024

A recent news release from Data storage company WEKA and S&P Global Market Intelligence unveiled the findings of their second annual Global Trends in AI report. The global study, conducted by S&P Global Market In Read more…

Argonne’s HPC/AI User Forum Wrap Up

September 11, 2024

As fans of this publication will already know, AI is everywhere. We hear about it in the news, at work, and in our daily lives. It’s such a revolutionary technology that even established events focusing on HPC specific Read more…

Quantum Software Specialist Q-CTRL Inks Deals with IBM, Rigetti, Oxford, and Diraq

September 10, 2024

Q-CTRL, the Australia-based start-up focusing on quantum infrastructure software, today announced that its performance-management software, Fire Opal, will be natively integrated into four of the world's most advanced qu Read more…

Computing-Driven Medicine: Sleeping Better with HPC

September 10, 2024

As a senior undergraduate student at Fisk University in Nashville, Tenn., Ifrah Khurram's calculus professor, Dr. Sanjukta Hota, encouraged her to apply for the Sustainable Research Pathways Program (SRP). SRP was create Read more…

GenAI: It’s Not the GPUs, It’s the Storage

September 12, 2024

A recent news release from Data storage company WEKA and S&P Global Market Intelligence unveiled the findings of their second annual Global Trends in AI rep Read more…

Shutterstock 793611091

Argonne’s HPC/AI User Forum Wrap Up

September 11, 2024

As fans of this publication will already know, AI is everywhere. We hear about it in the news, at work, and in our daily lives. It’s such a revolutionary tech Read more…

Quantum Software Specialist Q-CTRL Inks Deals with IBM, Rigetti, Oxford, and Diraq

September 10, 2024

Q-CTRL, the Australia-based start-up focusing on quantum infrastructure software, today announced that its performance-management software, Fire Opal, will be n Read more…

AWS’s High-performance Computing Unit Has a New Boss

September 10, 2024

Amazon Web Services (AWS) has a new leader to run its high-performance computing GTM operations. Thierry Pellegrino, who is well-known in the HPC community, has Read more…

NSF-Funded Data Fabric Takes Flight

September 5, 2024

The data fabric has emerged as an enterprise data management pattern for companies that struggle to provide large teams of users with access to well-managed, in Read more…

Shutterstock 1024337068

Researchers Benchmark Nvidia’s GH200 Supercomputing Chips

September 4, 2024

Nvidia is putting its GH200 chips in European supercomputers, and researchers are getting their hands on those systems and releasing research papers with perfor Read more…

Shutterstock 1897494979

What’s New with Chapel? Nine Questions for the Development Team

September 4, 2024

HPC news headlines often highlight the latest hardware speeds and feeds. While advances on the hardware front are important, improving the ability to write soft Read more…

Critics Slam Government on Compute Speeds in Regulations

September 3, 2024

Critics are accusing the U.S. and state governments of overreaching by including limits on compute speeds in regulations and laws, which they claim will limit i Read more…

Everyone Except Nvidia Forms Ultra Accelerator Link (UALink) Consortium

May 30, 2024

Consider the GPU. An island of SIMD greatness that makes light work of matrix math. Originally designed to rapidly paint dots on a computer monitor, it was then Read more…

AMD Clears Up Messy GPU Roadmap, Upgrades Chips Annually

June 3, 2024

In the world of AI, there's a desperate search for an alternative to Nvidia's GPUs, and AMD is stepping up to the plate. AMD detailed its updated GPU roadmap, w Read more…

Atos Outlines Plans to Get Acquired, and a Path Forward

May 21, 2024

Atos – via its subsidiary Eviden – is the second major supercomputer maker outside of HPE, while others have largely dropped out. The lack of integrators and Atos' financial turmoil have the HPC market worried. If Atos goes under, HPE will be the only major option for building large-scale systems. Read more…

Nvidia Shipped 3.76 Million Data-center GPUs in 2023, According to Study

June 10, 2024

Nvidia had an explosive 2023 in data-center GPU shipments, which totaled roughly 3.76 million units, according to a study conducted by semiconductor analyst fir Read more…

Shutterstock_1687123447

Nvidia Economics: Make $5-$7 for Every $1 Spent on GPUs

June 30, 2024

Nvidia is saying that companies could make $5 to $7 for every $1 invested in GPUs over a four-year period. Customers are investing billions in new Nvidia hardwa Read more…

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 1024337068

Researchers Benchmark Nvidia’s GH200 Supercomputing Chips

September 4, 2024

Nvidia is putting its GH200 chips in European supercomputers, and researchers are getting their hands on those systems and releasing research papers with perfor Read more…

Google Announces Sixth-generation AI Chip, a TPU Called Trillium

May 17, 2024

On Tuesday May 14th, Google announced its sixth-generation TPU (tensor processing unit) called Trillium.  The chip, essentially a TPU v6, is the company's l Read more…

Leading Solution Providers

Contributors

IonQ Plots Path to Commercial (Quantum) Advantage

July 2, 2024

IonQ, the trapped ion quantum computing specialist, delivered a progress report last week firming up 2024/25 product goals and reviewing its technology roadmap. Read more…

Intel’s Next-gen Falcon Shores Coming Out in Late 2025 

April 30, 2024

It's a long wait for customers hanging on for Intel's next-generation GPU, Falcon Shores, which will be released in late 2025.  "Then we have a rich, a very Read more…

xAI Colossus: The Elon Project

September 5, 2024

Elon Musk's xAI cluster, named Colossus (possibly after the 1970 movie about a massive computer that does not end well), has been brought online. Musk recently Read more…

Department of Justice Begins Antitrust Probe into Nvidia

August 9, 2024

After months of skyrocketing stock prices and unhinged optimism, Nvidia has run into a few snags – a  design flaw in one of its new chips and an antitrust pr Read more…

MLPerf Training 4.0 – Nvidia Still King; Power and LLM Fine Tuning Added

June 12, 2024

There are really two stories packaged in the most recent MLPerf  Training 4.0 results, released today. The first, of course, is the results. Nvidia (currently 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…

Spelunking the HPC and AI GPU Software Stacks

June 21, 2024

As AI continues to reach into every domain of life, the question remains as to what kind of software these tools will run on. The choice in software stacks – Read more…

Shutterstock 1886124835

Researchers Say Memory Bandwidth and NVLink Speeds in Hopper Not So Simple

July 15, 2024

Researchers measured the real-world bandwidth of Nvidia's Grace Hopper superchip, with the chip-to-chip interconnect results falling well short of theoretical c Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire