Duke University anticipates $100,000 or more per year in energy savings, while increasing processing power.

By Nicole Hemsoth

November 1, 2010

Duke University anticipates $100,000 or more per year in energy savings, while increasing processing power.

Crunching large quantities of data is a crucial component of modern academic life. On university campuses, everyone from chemists and engineers to researchers in sociology, economics and even literature needs access to high-performance computing (HPC). At Duke University, the Duke Shared Cluster Resource (DSCR) supports these needs for professors across the spectrum.

The DSCR serves 650 end users in 70 different research groups. One major research area that the DSCR caters to is the bioinformatics community. “Several of our professors are doing genome comparisons,” says John Pormann, Duke’s director of scalable computing. “Some researchers are building models that establish probabilities for when certain genes arose—and when species with a common ancestor became distinct. One is trying to develop the tree of life for a set of fungi by looking at genetics.”

Many bioinformatics researchers need to process an unprecedented volume of data. “One of our professors is working with a grant agency that gives grant recipients the newest, fastest gene-sequencing machine, rather than offering a large amount of money,” Pormann says. “Professors receive a cutting-edge gene sequencer, which spits out enormous amounts of data. They depend on the DSCR to provide an HPC environment that supports their data analysis needs.”

Cutting-edge research using limited power

In its efforts to meet those needs, the DSCR faces substantial logistical challenges. The building that houses the data center was constructed to hold classrooms, not a server farm. The data center has already maxed out the amount of electricity that can run through the building’s conduit. Its floor is raised only 12 inches, and the last air conditioning installation involved overhead units because the installers weren’t convinced they could push enough air through the floor to keep all the equipment cool.

“If we exceed the capacity of our current air conditioning unit, we have to go back to the drawing board,” says Pormann. “Adding power to our current location would be expensive because we’d have to bring in new power lines from the street, digging up sidewalks and parking lots in the process. Moving to a new data center would cost well into the six figures. We’d like to stay where we are, but that means we need to use the power and cooling in that room as efficiently as possible. It’s imperative as we expand our computing power that we keep our power and cooling resources constant.”

168 fewer watts per server

The DSCR needs to maximize the ratio of processing power to energy consumption for servers in its HPC cluster. The cluster encompasses 729 servers, from Dell PowerEdge M610 blade servers with Intel Xeon processor 5600 series to seven-year-old rack-mounted boxes. Oracle Grid Engine dynamically provisions application workloads to the most appropriate resources in the cluster.

To gain insight into the DSCR’s balance between processing power and consumption of energy and cooling resources, Pormann embarked on research of his own. He studied how CPU utilization correlates to power consumption in the cluster. For more than a year he collected data on the wattage drawn by the cluster’s primary servers, all Dell dual-CPU, quad-core machines, running their normal workload.

The comparison between machines at 100 percent CPU utilization is striking: Each of the data center’s rack-mounted Dell PowerEdge 1950 servers with Intel Xeon processor X5355 uses 369 watts. Its Dell PowerEdge M600 blade servers with Intel Xeon processor E5420 use 221 watts each. And its PowerEdge M610 blades with Intel Xeon processor E5520 use 201 watts. In three generations, PowerEdge servers reduced their power consumption by 168 watts, or 46 percent, at 100 percent CPU utilization.

“Across the board, from idle performance up to 100 percent CPU utilization, we see significant drops in power consumption on new Dell blade servers,” Pormann reports. “The form factor itself reduces energy needs because blades have fewer, larger, more efficient power supplies, and that’s clearly a factor in the sharp decline in power consumption that we see between the PowerEdge 1950 servers and the PowerEdge M610 blades.”

Comparing the Dell PowerEdge M600 blades with Intel Xeon processor E5420 and the PowerEdge M610 blades with Intel Xeon processor E5520 shows the difference a more efficient processor can make. “The across-the-board drop of roughly 20 watts is primarily attributable to CPU improvements,” says Pormann. “This suggests that all workloads would see significant improvements in energy savings from simply moving to the new CPUs. No matter what load is placed on the system, we should see significant reductions in power usage.”

$100,000-plus per year in energy cost savings

In addition to helping the DSCR stay in its current location, reducing servers’ power usage—both directly and by reducing air conditioning usage—saves the university money on an ongoing basis. “Even with North Carolina’s below-average power costs, we’re seeing savings of $100,000 or more per year, just by reducing our energy usage,” Pormann says.

As energy consumption falls, processing power grows dramatically with each new generation of Dell blades. “Every time we add a Dell PowerEdge M610 blade, we can pull older nodes from the cluster and retain the same processing power,” Pormann says. “When we buy new equipment, our users are always astounded at how much faster their jobs run. But we can’t directly calculate how many older nodes are comparable to a new blade. For one researcher, a new blade might replace eight older servers, but for another researcher that ratio might be 12 to 1.”

The DSCR just implemented six new 12-CPU-core PowerEdge M610 blades with Intel Xeon processor 5600 series and 96 gigabytes of memory per blade. “We’re really looking forward to seeing what they can do,” says Pormann. “Our early data indicates that they use no more power than our 8-CPU-core PowerEdge M610 blades with Intel Xeon processor E5520. So we should see a 50 percent increase in computational capability with zero increase in power consumption.”

Available capacity in the HPC cluster indicates that the new Dell PowerEdge blades are giving the cluster a substantial boost in processing power. Historically, the DSCR has run at about 70 percent of its total capacity. Since recent additions of Dell blades, utilization has declined to around 50 percent, so the cluster’s unused capacity has increased by as much as 30 percent. “Because we have more head room, we have the capability for researchers to do different kinds of experimentation,” says Pormann. “They don’t have to worry about wasting capacity anymore. They appreciate that whenever they need more computational power, it’s here.”

Pormann attributes the performance gains of the latest Dell servers to advancements, in part, in their management of memory. “More and more of our users are asking about large memory configurations,” he says. “It seems research projects are starting to be constrained by node memory performance and capacity. Over the last nine months, three different professors involved in bioinformatics have asked me whether they can get 256 or 512 gigabytes of memory in a single blade. One of them is working with images, each of which is on the order of a terabyte in size. I told them that systems with those memory configurations are on the Dell road map. The integrated memory controller and the Intel QuickPath architecture give the Dell blades exceptional bandwidth for memory processing.”

The DSCR further improves performance within the cluster using the Intel Compiler Suite. “We’re leveraging the Intel compilers as much as possible,” Pormann says. “What we’ve seen so far is that the Intel compilers, compared with the open source compilers, provide anywhere from 20 to 50 percent improvement in performance. You run the exact same C code through the Intel compiler, and the executable is faster. The Intel compilers seem to take advantage of all the bells and whistles in these new Intel processors.”

Server management tools bring additional efficiencies

Now Pormann is figuring out better ways to use the tools at his disposal to manage power for the cluster’s servers. The DSCR uses Dell Chassis Management Controller (CMC) to monitor energy consumption of the blades. Then IPMItool, an open source utility, exports this data to the Oracle Grid Engine. “Dell has its own add-ons to IPMItool,” Pormann says. “We were able to talk to the Dell engineering group, and they gave us information about the command line interfaces we could use to gather this data for Grid Engine.”

Pormann and his team currently can use Dell CMC to remotely control power usage within the cluster. “Dell CMC should enable us to easily power off machines and idle groups of machines to reduce the cluster’s overall heat load,” says Pormann.

The next step is to automate the process of powering off servers through Oracle Grid Engine. “An idle machine still uses more than 100 watts,” says Pormann. “Throttling the CPU may reduce power consumption, but it should be used in conjunction with powering off unused machines. And our systems should be capable of powering machines down automatically so someone doesn’t have to push buttons 24×7.  Knowing we have this potential in the blades helps us justify our continuing commitment to building out these tools.”

Mapping the cluster’s future

Duke purchases most of the cluster’s hardware from Dell for several reasons. One is the product road map from Dell and Intel. “We like Dell’s track record of always putting the newest equipment from Intel in their hardware,” Pormann says. “When Intel announces a new chip, we know we’re going to see it in Dell’s equipment shortly. Knowing what’s coming on those road maps enables us to spend our money very wisely.”

Perhaps the most important reason why the DSCR continues to purchase from Dell is that Pormann has always been pleased with the support he’s received. “Anytime we have a hardware issue, Dell cross-ships us parts and gives us access to engineers,” says Pormann. “As long as Dell and Intel continue to work on efficiency in the ratio of processing power to energy consumption, we’ll be able to continue to increase computational power, even as we stay in our current data center.”

For more information visit Dell.

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