Optimization of Data Center Cooling and Water Efficiency Part 1: Series Introduction

By Chris Reininger

Keeping servers cool is one of the biggest challenges data centers face due to the rapid growth of AI and its energy-intensive graphics processing units (GPUs). These high-density workloads generate massive heat, which must be rejected to keep equipment functioning properly. Data centers typically use a combination of air and water to transfer and reject heat to the atmosphere. These processes not only consume huge amounts of energy but also billions of gallons of water, up to 5 million gallons per day in large data centers.

In this series, we discuss these challenges and how to minimize energy and water consumption using both established and emerging technologies, such as immersion cooling.

We will cover:

• Heat rejection strategies and optimization, including evaporative cooling towers, dry coolers, closed circuit cooling, hybrid/adiabatic coolers. All these strategies can be made more efficient with proper application, adherence to ASHRAE’s latest recommendations for data center cooling, and careful equipment selection and system control. We will discuss these best practices along the way.

• Utilizing waste heat, how to recover heat and where to use it.

• Minimizing water use within a given heat rejection strategy, as well as rainwater harvesting, gray water collection, and the associated filtration and treatment required.

• Emergency power systems and generators, with an emphasis on efficient exhaust handling.

It’s a Balancing Act

One caveat that makes cooling modern data centers so difficult is that some high-efficiency strategies use more water, depending on the heat-rejection methods chosen. For instance, evaporative cooling is very efficient; however, a large, evaporative-cooled data center can consume up to 1.8 billion gallons of water annually. Data centers will often combine strategies to avoid these kinds of trade-offs.

Data center cooling design is a balancing act. This is especially true when it comes to heat rejection. Many factors must be weighed against others, including the size of the data center, rack density, location and climate, demand load profile, water source, the building's physical layout and structure, reliability, and redundancy requirements, and more.

This series explores that balancing act, comparing the pros and cons of various types of heat rejection and how to mitigate inefficiencies.

Next up: Data Center Heat Rejection