Which Type of Data Center Cooling System Is Best for You?

If data is the world’s most valuable resource, data centers are among its most critical facilities. They’re also some of the fastest-growing operations, a trend that will only increase as demands rise. Businesses must consider how to build and operate them more efficiently, and cooling is a big part of that equation.

Data center cooling is crucial to keeping servers running efficiently. It’s also the most energy-intensive part of running one of these facilities. Consequently, businesses wanting to make their data centers as cost-efficient and functional as possible must consider what kind of system they use.

Types of Data Center Cooling Systems by Placement

Choosing a data center cooling system starts with knowing what’s available. There are a few ways to distinguish between different methods, with one of the most common being where the cooling occurs.

Room Cooling

Room cooling — also called perimeter cooling — is the most straightforward data center cooling method. This approach places air conditioning units around the server room’s perimeter, much like many home AC systems work.

Perimeter cooling’s simplicity is its biggest advantage. It’s easy to implement and has relatively low upfront costs. However, it uses a lot of energy. Considering how data centers already consume up to 50 times more power than other buildings, that’s not ideal.

Because the hardware necessary for this method is so affordable, it’s easy to scale or adjust room cooling systems. However, this scalability is limited. Once data centers reach a certain density, room cooling may not effectively cool all the servers, leading to diminishing returns.

Row Cooling

Row cooling is a more efficient alternative. This setup places cooling systems along each row of servers in the data center. In some cases, in-row systems place units within the rows themselves, providing more direct airflow.

In-row cooling costs more to implement than perimeter cooling but is more energy efficient. It’s also more targeted, which can lower fan power consumption by more than 50% by eliminating wasted airflow. Targeted cooling also lets businesses direct more resources to specific rows to improve reliability for mission-critical servers.

Despite those advantages, row cooling isn’t perfect in every scenario. Some data centers can’t accommodate rows, and others have too many mission-critical servers to direct more airflow to one row than others. Row cooling can also be challenging to implement in non-rectangular rooms.

In-Rack Cooling

In-rack cooling is the natural next step from row cooling. It places units in each server rack, providing the most direct temperature control of the three major system types.

In-rack cooling is the most energy-efficient method, making it more environmentally friendly and the cheapest to operate in terms of power costs. It’s easy to scale, too, as installing a cooling unit on an existing server rack is relatively straightforward. This method offers even more targeted cooling than in-row techniques.

The biggest downside to in-rack cooling is its high upfront costs. While its targeted nature is good for system performance, it doesn’t offer much redundancy, which can be an issue. If one unit fails, there’s no roomwide cooling system to lower temperatures and make up for it.

Types of Data Center Cooling Systems by Technology

Within these location-based categories, data center cooling systems can vary by technology. Here are some of the most popular types.

Air Cooling

The most straightforward way to cool data center servers is with air coolers. It’s the most common cooling system in personal computers and is equally popular in data centers for many of the same reasons. Blowing cold air over hot components is relatively inexpensive, easy and reliable.

Computer room air conditioners (CRACs) are the most popular air coolers and work like conventional AC units. Some data centers use computer room air handlers (CRAHs) instead. These use cold water to cool outside air, whereas air conditioners use chemical refrigerants to cool inside air. CRAHs are more expensive upfront but use less energy.

All air cooling systems are relatively simple and easy to implement but aren’t as efficient as other options. They can also be loud and bulky, limiting their usefulness in data centers with limited space.

Liquid Cooling

Liquid cooling is a popular alternative to air-based systems. These systems use liquids — either water or a refrigerant — to pass by server components, absorbing their heat and transferring it away for the part to cool. Because water’s thermal conductivity is 25 times higher than air’s, liquid cooling is much more efficient.

There are two main types of liquid cooling for data centers — immersion and direct-to-chip. Direct-to-chip cooling uses tubes to run coolants over the warmest components of a server and carry it away, like how most liquid-cooled PCs work. Immersion is a more extreme technique that places the entire server in a container of nonconductive cooling liquid.

All types of liquid cooling are more efficient than air cooling, with immersion offering the most efficiency because it involves no active parts. However, they’re also more expensive and challenging to implement. Because electronics and liquids generally don’t mesh well, any errors can be catastrophic.

Geothermal Cooling

Some newer data centers use a completely different method — geothermal cooling. This practice uses the natural lower temperatures of the earth below ground to cool servers. Geothermal cooling often uses heat pumps similar to large-scale liquid cooling systems to transfer underground temperatures to above-ground data centers.

Because geothermal cooling capitalizes on constant natural temperatures, it’s remarkably efficient. One data center in Pennsylvania uses 34% less energy per year after implementing a geothermal solution. The system also reduces maintenance costs and offers more reliability by minimizing moving parts.

Despite its environmental and reliability benefits, geothermal cooling is expensive. Some sites may also have geographic limitations or zoning restrictions that stop them from using it.

How to Choose a Data Center Cooling System

Choosing the right data center cooling system starts with understanding what a business needs. Reliability may be more important than costs if the servers handle lots of sensitive data. In that case, liquid cooling and row or in-rack systems are better than air and room cooling solutions.

Businesses should review their regulatory requirements to gauge which factors are most important in their cooling. SOC2 compliance requires proof of servers’ availability and processing integrity, which means reliability and redundancy are crucial in a cooling system. Alternatively, companies that face strict environmental regulations should prioritize energy efficiency.

Generally speaking, geothermal cooling is the most energy-efficient, but liquid cooling is more efficient than air-based systems. However, these methods cost more to implement, so they’re not ideal for every budget. Smaller companies with less environmental pressure can use air cooling or a more efficient setup like in-rack cooling to offset their energy consumption.

It’s also important to consider the data center’s size. Room cooling may be sufficient for small facilities, but larger or more dense operations need more targeted systems. Sometimes, it’s best to combine solutions for redundancy. For example, a data center could use liquid cooling with a room-cooled air solution as a backup.

Data Centers Have Different Cooling Needs

There’s no one answer to what the best data center cooling system is. Businesses must look at what’s available and compare it to their needs to see what works best for them.

New cooling methods will emerge as technology improves. Keeping up with these changes will help companies make the most informed decisions about balancing efficiency, costs, reliability and sustainability. Doing that can create an optimal data center for their unique situation.