Earth Day 2020 arrives at a strangely terrible and terribly strange juncture of events. With COVID-19 shuttering businesses and stunning large swathes of global economies, the internet, service providers and cloud computing platforms are playing key roles in everything from streaming entertainment to social networking to educational programming to enabling thousands of companies’ work from home (WFH) policies.
In addition, high-performance systems and supercomputers are being used to explore key issues around the pandemic and chart the path to a vaccine.
But at the same time, large-scale cloud data centers and supercomputing installations are among the world’s biggest consumers of electrical power, much of it produced by means of unsustainable fossil fuel resources. Are IT vendors and their customers doing anything to address these issues? Actually, yes, with efforts that seem appropriate to discuss today, Earth Day 2020.
Bringing some light to the cloud
Public cloud platforms have often claimed high ground in vocally using and supporting sustainable energy resources and projects. Those efforts include developing data centers near renewable power sources: Google’s data center at The Dalles, Ore. and the facilities that Facebook and Apple constructed in Prineville, Ore. are good examples. In fact, four companies with large public cloud operations were highlighted in the 2019 Top 10 buyers list compiled by the Renewable Energy Buyers Alliance (REBA).
However, the volume of renewable energy purchased only tells one part of a broader story. In December 2019, Wired magazine looked into the renewable energy programs at Google Cloud, Microsoft Azure and AWS. The details it revealed were intriguing and served as a good reminder of how difficult and complicated a dedication to sustainable resources can be, especially for large and ambitious cloud vendors.
Sustainable energy usage can also be an issue when it comes to high performance and supercomputing, at least in regard to individual systems. Many of the world’s largest supercomputing installations reside in facilities that pre-date those systems. For example, the U.S. Department of Energy operates laboratories in Oak Ridge, Tenn., Los Alamos, N.M. and Livermore, Calif. Not surprisingly, access to renewable energy resources in those locations is variable.
However, systems can also be designed and situated to take advantage of available sustainable resources. In 2019, IBM donated an $11.6 million computer cluster to the Massachusetts Institute of Technology featuring the architecture of the company used in the Summit supercomputer it built at the DOE’s Oak Ridge Lab (cited as the world’s fastest supercomputer in the November 2019 Top500.org list).
Named Satori, a Zen Buddhism term for “sudden enlightenment,” the MIT cluster is the size of a shipping container and supports up to 2 quadrillion calculations per second (2 petaFLOPS). However, the carbon footprint of the system’s 150 kilowatts power consumption will be nearly fully offset by the use of carbon emissions-free energy sources.
Why sustainable tech is critical for AI
Not surprisingly, Satori was used in January to host a three-day Green AI Hackathon sponsored by MIT and IBM. But why is artificial intelligence a critical subject in terms of sustainable energy? Because while many consider AI to be one of the most promising new commercial opportunities on the IT landscape, it is also one of the most energy intensive. In fact, a 2018 study cited by OpenAI estimates that the amount of compute required for AI training runs has grown exponentially, doubling every 3.4 months since 2012 or more than 300,000x during that time.
AI-enabled workloads have also become commonplace in some advanced supercomputers, enabling speedier analysis of massively complex datasets and computations. That includes the DOE’s Summit, Sierra and Lassen supercomputers that IBM built using a hybrid architecture (leveraging its own POWER9 CPUs, NVIDIA GPUs and Mellanox Infiniband). Those systems were respectively ranked #1, #1 and #10 on the November 2019 top500.org list.
But IBM’s installations are also notably energy efficient, with Summit ranking #5 and Sierra ranking #10 on the November 2019 Green500 list from Top500.org. Plus, a third IBM hybrid system—the PANGEA III owned by Total Exploration Production—earned the #9 Green500 rank on that list. How do these systems deliver both world-class compute and “green” performance? Via design innovations and system management features, including a self-regulated water-cooling system designed to analyze water temperature fluctuations and predict the optimal settings for enhanced energy efficiency.
Computing technologies play critical roles in a wide variety of processes and practices. Cloud vendors are instrumental in supporting crucial consumer and business services. Research facilities worldwide are lending and employing their supercomputing resources to research efforts around COVID-19, along with projects concerning the effects of climate change and global warming.
While concerns about the energy requirements, consumption and demand of large-scale computing facilities remain on the table, it is also clear that innovative vendors, service providers and researchers are determined to seek and find workable, sustainable solutions to those problems. That is a point worth recognizing and celebrating on Earth Day 2020.
Charles King is a principal analyst at PUND-IT and a regular contributor to eWEEK. © 2019 Pund-IT, Inc. All rights reserved.