SC13 Research Highlight: COCA Targets Datacenter Costs, Carbon Neutrality

By Shaolei Ren and Yuxiong He

November 16, 2013

The rapid growth of high performance computing and cloud computing services in recent years has contributed to the dramatic increase in the number and scale of data centers, resulting in a huge demand for electricity.

According to recent studies, the combined electricity consumption of global data centers amounts to 623 billion kWh annually and would rank 5th in the world if the data center were a country. As a significant portion of electricity is produced by coal or other carbon-intensive sources, it is often labeled as “brown energy” and the growing trend of the data center electricity consumption has raised serious concerns about its carbon footprint as well as environmental impacts, such as altering global patterns of temperature, rainfall, and creation of drought and flood.

Recently, large data center operators such as Google and Microsoft have been increasingly urged to find effective solutions to reduce their carbon emissions for sustainable computing and ultimately achieve an overall net-zero carbon footprint (i.e., carbon neutrality), as mandated by governments in the form of Kyoto-style protocols, voluntarily for public images, or urged by environmental organizations.

While it is beneficial for sustainability, achieving carbon neutrality presents significant challenges for data center operators, because the best location for building a data center may not be the most desired location for generating sufficient green energies (e.g., solar, wind) to satisfy the data center requirement and using carbon-free electricity directly from utility companies is not yet widely available. Thus, as a practical alternative, carbon-neutral data centers often rely on a bundle of approaches, such as generating off-site green (or renewable) energy and purchasing renewable energy credits (RECs): using renewable energy to indirectly offset electricity usage.

Completely offsetting electricity usage via off-site renewable energy generation for long-term carbon neutrality is desirable yet challenging: data centers need to carefully budget electricity usage over a long timescale (often a year) such that the “unknown” future brown energy consumption can be completely offset by limited renewables. While it seems to be easy to plan the electricity usage over a long timescale based on the future computing demand, a practical challenge is that the far future time-varying workloads or intermittent renewable energy availability cannot be accurately predicted and hence data centers need to decide electricity usage in an online manner.

In our research, we study long-term energy budgeting for a carbon-neutral data center and propose a provably-efficient online resource management algorithm, called COCA (optimizing for COst minimization and CArbon neutrality), for minimizing the operational cost while satisfying carbon neutrality without requiring long-term future information. Both electricity cost and delay performance are incorporated into our optimization objective.

COCA eliminates the requirement of knowing long-term future computing demand information by keeping track of the “carbon deficit” online. As the name implies, carbon deficit indicates how far the current data center operation deviates from carbon neutrality, or more precisely, how much the current electricity usage has exceeded the available renewables. Incorporating the carbon deficit into the optimization objective, COCA progressively adheres to carbon neutrality by adapting the weight of electricity consumption. Specifically, if the carbon deficit is larger, COCA will place more emphasis on reducing electricity consumption by turning down more servers such that the carbon deficit can be offset by future renewables. Thus, COCA works following the philosophy of “if violating carbon neutrality, then use less electricity”.  While the intuition is straightforward, we formally prove by extending the recently-developed Lyapunov optimization that COCA achieves a close-to-minimum operational cost, compared to the optimal offline algorithm with look-ahead information, while bounding the maximum possible carbon deficit.

Large data centers often consist of up to tens of thousands of servers, and distributed server management is highly desirable for scalability. Towards this end, we embed distributed resource management in COCA such that each server autonomously adjusts its processing speed (and hence, power consumption, too) and optimally decides the amount of workloads to process. Specifically, each server can “learn” the optimal decision by sampling a set of possible decisions and eventually choosing the best one with very high probability.

To validate COCA, we perform an extensive simulation study modeling the one-year operation of a large data center. Using real-world production traces to drive the simulation, we first compare COCA against state-of-the-art prediction-based method in terms of the average hourly operational cost. In particular, PerfectHP (Perfect Hourly Prediction heuristic), which perfectly predicts 48-hour-ahead workloads and allocates the carbon budget in proportion to the hourly workloads, is chosen as the benchmark. As shown in the figure, COCA is more cost effective compared to PerfectHP with a cost saving of more than 25% over one year. COCA achieves the benefit because even though the workload spikes and carbon neutrality is temporarily violated, it can focus on cost minimization while carbon deficit will then later guide the data center operation towards carbon neutrality. By contrast, without foreseeing the long-term future, short-term prediction-based PerfectHP may over-allocate the carbon budget at inappropriate time slots and thus have to set a stringent budget for certain time slots when the workload is high.


Next, we show that, under different electricity usage budgets, the operational cost of COCA is always fairly close to the minimum value achieved by the optimal offline algorithm with complete future information. Operational cost and electricity usage are normalized with respect to the carbon-unaware algorithm that disregards carbon neutrality and purely minimizes the operational cost. It can be seen that with a normalized electricity usage of 0.9 (i.e., saving 10% electricity usage compared to the carbon-unaware algorithm), COCA only increases the operational cost by less than 3% compared to both the carbon-unaware algorithm and the optimal offline algorithm that has the complete future information. This demonstrates a strong applicability of COCA in real systems due to its good performance and online execution without complete offline information.


To summarize, COCA addresses carbon neutrality, an emerging issue in data centers: it enables data centers to achieve a low operational cost while satisfying carbon neutrality in the absence of long-term future information. The distinguishing feature of online and distributed implementation makes COCA an appealing candidate for autonomously managing computing resources in large data centers.

SESSION: Performance Management of HPC Systems


TIME: Tuesday, 11:30AM – 12:00PM

ROOM: 401/402/403

Shaolei Ren received his Ph.D. from University of California, Los Angeles, in 2012 and is currently with Florida International University as an Assistant Professor. His research focuses on sustainability and emerging topics in cloud computing such as water usage effectiveness.

Yuxiong He is a researcher at Microsoft Research.  Her research interests include resource management, algorithms, modeling and performance evaluation of parallel and distributed systems.  Her recent work focuses on improving responsiveness, quality and throughput of large-scale interactive cloud services such as web search.  Yuxiong received her Ph.D. in Computer Science from Singapore-MIT Alliance in 2008.

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!

Aurora AI-Driven Atmosphere Model is 5,000x Faster Than Traditional Systems

July 16, 2024

While the onset of human-driven climate change brings with it many horrors, the increase in the frequency and strength of storms poses an enormous threat to communities across the globe. As climate change is warming ocea Read more…

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 claims. A paper published on July 10 by researchers in the U. Read more…

Belt-Tightening in Store for Most Federal FY25 Science Budets

July 15, 2024

If it’s summer, it’s federal budgeting time, not to mention an election year as well. There’s an excellent summary of the curent state of FY25 efforts reported in AIP’s policy FYI: Science Policy News. Belt-tight Read more…

Peter Shor Wins IEEE 2025 Shannon Award

July 15, 2024

Peter Shor, the MIT mathematician whose ‘Shor’s algorithm’ sent shivers of fear through the encryption community and helped galvanize ongoing efforts to build quantum computers, has been named the 2025 winner of th Read more…

Weekly Wire Roundup: July 8-July 12, 2024

July 12, 2024

HPC news can get pretty sleepy in June and July, but this week saw a bump in activity midweek as Americans realized they still had work to do after the previous holiday weekend. The world outside the United States also s Read more…

Nvidia, Intel not Welcomed in New Apple AI and HPC Development Tools

July 12, 2024

New Mac developer tools will leverage Apple's homegrown chips, limiting HPC users' ability to use parallel programming frameworks from Intel or Nvidia. Apple's latest programming framework, Xcode 16, was introduced at Read more…

Aurora AI-Driven Atmosphere Model is 5,000x Faster Than Traditional Systems

July 16, 2024

While the onset of human-driven climate change brings with it many horrors, the increase in the frequency and strength of storms poses an enormous threat to com 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…

Shutterstock 2203611339

NSF Issues Next Solicitation and More Detail on National Quantum Virtual Laboratory

July 10, 2024

After percolating for roughly a year, NSF has issued the next solicitation for the National Quantum Virtual Lab program — this one focused on design and imple Read more…

NCSA’s SEAS Team Keeps APACE of AlphaFold2

July 9, 2024

High-performance computing (HPC) can often be challenging for researchers to use because it requires expertise in working with large datasets, scaling the softw Read more…

Anders Jensen on Europe’s Plan for AI-optimized Supercomputers, Welcoming the UK, and More

July 8, 2024

The recent ISC24 conference in Hamburg showcased LUMI and other leadership-class supercomputers co-funded by the EuroHPC Joint Undertaking (JU), including three Read more…

Generative AI to Account for 1.5% of World’s Power Consumption by 2029

July 8, 2024

Generative AI will take on a larger chunk of the world's power consumption to keep up with the hefty hardware requirements to run applications. "AI chips repres Read more…

US Senators Propose $32 Billion in Annual AI Spending, but Critics Remain Unconvinced

July 5, 2024

Senate leader, Chuck Schumer, and three colleagues want the US government to spend at least $32 billion annually by 2026 for non-defense related AI systems.  T Read more…

Point and Click HPC: High-Performance Desktops

July 3, 2024

Recently, an interesting paper appeared on Arvix called Use Cases for High-Performance Research Desktops. To be clear, the term desktop in this context does not 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…

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…

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…


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…

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…

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…

Some Reasons Why Aurora Didn’t Take First Place in the Top500 List

May 15, 2024

The makers of the Aurora supercomputer, which is housed at the Argonne National Laboratory, gave some reasons why the system didn't make the top spot on the Top 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…

Leading Solution Providers


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…

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…

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…

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…

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…

The NASA Black Hole Plunge

May 7, 2024

We have all thought about it. No one has done it, but now, thanks to HPC, we see what it looks like. Hold on to your feet because NASA has released videos of wh 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…

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…

  • arrow
  • Click Here for More Headlines
  • arrow