Recent computing advancements are forcing the data center industry to examine some harsh truths regarding power use.

The International Energy Agency (IEA) recently revealed that U.S. data centers consumed over 4 percent of the country’s total electricity in 2022, a figure anticipated to rise to 6 percent by 2026. Electricity demand in the United States, stagnant for two decades, showed a noticeable surge in the last year, and electric utilities have nearly doubled their forecasts of how much additional power they’ll need by 2028.

A major culprit is the unprecedented boom in cloud and AI applications, alongside an increasing number of data centers being constructed to accommodate growing demands. At the same time, local communities are increasingly reluctant to green-light data centers because of the limited power available. The inescapable fact is that the data center industry is energy-intensive, and the luxury of ample power access appears increasingly uncertain.

Generating more power through renewable sources and modernizing the grid are crucial steps toward addressing the power availability problem, but these solutions will take time to come to fruition. Furthermore, increasing power generation is not the same as increasing data center efficiency. Creating a more sustainable future for the data center industry is contingent on improving efficiency. We can do this by addressing the issue of stranded capacity, that is, capacity that is available in data centers but is going to waste. Stranded capacity is an issue we can solve here and now to immediately improve data center power efficiency.

Unlocking the hidden potential of stranded capacity

Addressing stranded capacity is crucial for maximizing the operational efficiency and sustainability of data centers. Stranded capacity in data centers refers to resources that are available but not used effectively due to power, cooling, or space constraints. This underutilization means that data centers are considered “full” well before they’ve reached maximum capacity.

Data center performance hinges on balancing capacity, efficiency, and compliance, which makes integrating evolving IT into fixed infrastructures especially challenging. As data centers update and remove equipment over time, power distribution becomes uneven, complicating airflow and cooling. The real challenge lies in the gap between the ideal and the real. Data center designs establish peak performance baselines, often referred to as “design performance.” However, operational data centers inevitably deviate from these configurations over time. As these deviations accumulate, the data center’s actual performance diverges from the optimal performance outlined in the design. Since design configurations represent peak potential, this means real-world operational performance will always be less than optimal. Deviations from the design configuration fragment resources, such as space, power, cooling, and airflow. This fragmentation often leads to overcompensation of resources to protect IT systems.

Current data center planning tools and techniques are hampered by a lack of visibility into resource fragmentation. This significantly hinders the data center management teams’ ability to maintain performance levels close to the original design. Effective management requires a deep understanding of available resources, necessitating a unified platform for tracking and planning across power, space, cooling, weight, and network capacities. However, many current management tools lack the predictive capabilities to anticipate the impacts of changes, such as adding high-density hardware, on power and cooling systems.

Accessing performance gains through critical insights

At Cadence, we’re pioneering the use of physics-based digital twins for data center management. Digital twins of data centers are essentially digital replicas of real-world data centers, built using physics-based simulations. These simulations, powered by techniques like computational fluid dynamics (CFD), mimic the physical behavior of their real-world counterparts. This powerful ability enables data center management teams to predict the impact of operational changes on performance. They can also use digital twins to explore different scenarios and optimize their plans for the best possible outcomes.

A model from Cadence Reality DC Design with a results plane showing the temperatures in the room and within the cold aisle containment.

The digital twin model stays current by incorporating data from existing operational tools like DCIM, environmental monitoring systems, and CSV spreadsheets. These integrations provide a super monitoring capability by combining a view of physical and virtual monitoring data and an always-ready digital twin model that can simulate capacity plans, conduct sensitivity studies, and optimize energy efficiency.

“Cadence’s state-of-the-art digital twin technology enables new benchmarks for energy efficiency in data centers,” said Marco Chiappetta, principal analyst at HotTech. “By incorporating high-fidelity computational fluid dynamics, Cadence empowers operators to achieve significant energy savings and optimal high-density AI deployments. This approach not only addresses the immediate efficiency demands of today but also offers a more sustainable path forward for AI data center management tomorrow.”

Revolutionizing data center management with CFD-powered digital twins

While the rapid growth of data centers and their escalating power demands are inevitable, using existing resources wisely can help manage and even significantly reduce power demand. Explore the Cadence data center solutions page to learn more about performance-aware design and operational planning of data centers through CFD-powered digital twins.

By: DocMemory
Copyright © 2023 CST, Inc. All Rights Reserved