Data center cooling is a hugely important issue, which is why it’s so commonly talked about amongst industry professionals.
It’s no secret that datacenters use a vast amount of electricity. Therefore, they often generate incredible levels of heat. This heat must be carefully controlled, as it poses serious levels of risk to the servers, storage devices and hardware housed within the datacenter.
Unfortunately, controlling the temperature of a datacenter isn’t as easy as it sounds. It can be tricky, expensive and it’ll often involve the consumption of huge amounts of energy to keep equipment housed in perfect conditions. In fact, cooling requirements are one of the key reasons why businesses tend to move to colocation datacenters, as in-house cooling can become inefficient in terms of both time and money.
Modern datacenters rely on a range of different cooling technologies to maintain optimal temperatures, reducing the risk of damage, data loss and downtime that can be caused by excessive heat. Read on and we’ll explain more about what datacenter cooling tech does, and why it’s so vital for growing businesses today.
What is Data Center Cooling?
Data center cooling refers to the various methods and technologies used to control the temperature in data centers, which are facilities that house computer systems and associated components such as telecommunications and storage systems.
These cooling systems are essential for maintaining the optimal performance and longevity of the equipment, as electronic devices generate a significant amount of heat during operation.
Why Does Datacenter Cooling Matter?
Cooling is one of the main priorities of datacenter management. Datacenters must be cooled at all times, in order to keep rising heat in check and protect the servers and hardware that datacenters have within them.
Storage devices, networking hardware and servers all generate heat to some level. When they’re working at full capacity, and surrounded by other servers and hardware doing exactly the same thing, heat levels can quickly start to increase beyond the temperatures that can be tolerated by this sensitive equipment. If the space isn’t then cooled, then damage can start to occur.
The main risks are damage to the fragile components of servers and hardware, and a reduction in the overall lifespan of different pieces of equipment. You don’t need us to tell you that much of this equipment represents a huge level of investment for a company, so any reduction in its lifespan will lead to considerable cost increases.
Datacenters are always kept cool to limit the risk of damage to components, and keep equipment working at full potential. This increases the lifespan of equipment and ensures that unexpected downtime can’t occur as a result of heat damage within the datacenter.
Current Datacenter Cooling Technology
Leading datacenters rely on a whole host of different cooling technologies to maintain specified temperatures. Take a look at some of the most commonly used cooling options in top datacenters today.
Computer Room Air Conditioner (CRAC)
A CRAC, or computer room air conditioner, is a type of air conditioner that has been specifically designed for use within datacenters and server rooms. They’re an affordable option, which is why we so often see them being used, but they do use a high level of energy and can therefore be expensive to run.
Computer Room Air Handler (CRAH): If a datacenter is located in a place with a relatively cool climate, a CRAH is a great option. It uses air from outside the datacenter’s walls, along with fans and chilled water to reduce the temperature inside.
Calibrated Vectored Cooling (CVC)
High density services are often cooled using a method called Calibrated Vectored Cooling, or CVC. This option makes it possible to increase the number of circuit boards on each server chassis, by improving airflow within the room.
Evaporative Cooling: Warm air from within the datacenter can be cooled using water, which will evaporate and cool the air down as a result. This is a good option if energy usage is a concern, but it does involve a high level of water consumption to function.
Free Cooling
Another energy efficient option that is great for datacenters that are located in cooler places, free cooling can make use of cold air from outside. It brings in colder air from outside the building to reduce the air temperature inside, rather than relying solely on technology that cools the temperature of the same air within the datacenter. Unfortunately, it won’t work in warmer climates, but it is brilliantly useful for many datacenters around the world.
Chilled Water System
Datacenters sometimes have their own chiller plants, which can be used to facilitate cooling via a chilled water system. This method uses chilled water to reduce the temperature in the data center, cooling the air using specially designed air handlers.
Cold Aisle and Hot Aisle Containment
Some datacenters rely on a method of cooling known as cold and hot aisle containment. This involves creating alternate cold and hot aisles, with air intakes and air exhausts on the front and back of racks. Hot air being expelled will therefore go straight into the air intakes to be cooled, and the overall temperature is reduced efficiently.
Raised Floor
Purpose-built datacenters can be designed with facilities for cooling technologies already built in. The creation of a raised floor is a prime example of this. It involves lifting the floor of the datacenter, to allow room for cooling pipes or better airflow beneath the floor. The space can also be used to run power cables through. When datacenters are designed in this way, they can be cooled more efficiently.
The Three Levels of Data Center Cooling
There are an incredible number of different options when it comes to data center cooling. Some levels, such as system-level cooling, are highly energy efficient, while others will use far more power, but provide much faster and more reliable cooling as a result.
Let’s take a look at the three main levels of data center cooling to explore how data centers are making use of technology to suit the specific requirements of their equipment.
Board-level cooling
Board-level components are responsible for much of the heat that data centers generate. Units like CPUs, GPUs and TPus generate incredible amounts of heat, and as the complexity of these units grows so too does the amount of heat they produce. Effective heat dissipation is therefore vital to keep the internal temperatures of units like these stable, and prevent failure.
In board-level cooling, air is used to absorb excess heat. Air-based cooling devices like heat sinks, heat pipes, vapor chambers, active cooling fans and blowers dissipate heat. In particularly high density data centers, liquid cooling systems are also often used. Usually, this is because the amount of heat being generated by the hardware here is far more than what can be cooled using more traditional methods.
Often liquid cooling is used in high-performance computing (think the training of AI applications and complex data analysis).
Sub-system cooling
While liquid cooling is incredibly effective, it is rarely used to cool entire data centers down. And that’s largely down to the cost factor. Liquid cooling can be prohibitively expensive, so it’s only really used when it’s absolutely necessary. Often, you’ll find data centers using a number of different cooling techniques in different areas, which is known as sub-system cooling.
Liquid cooling systems used in specific areas typically collect heat and dispose of it via liquid, which goes into the area. Traditional air-based cooling systems then complete the work, saving on cost.
Airflow is a real concern in the management of temperatures in data centers. Rack-level cooling systems are often used to direct airflow where it’s needed most, and keep air moving through the data center to avoid any areas growing too hot.
Walk through a data center and you’ll likely find areas organized into hot and cold aisles. This allows for air to circulate, and fans to take air from cold aisles to be used in the cooling of hot aisles. Sometimes physical barriers are also used to improve the efficiency of such setups.
System-level cooling
Some data centers are cooled on a system-level. This tends to happen in newer facilities, where it’s been possible to plan infrastructure around cooling needs and energy consumption. Typically, such data centers use a combination of different cooling systems, depending on the particular requirements of the facility and the equipment housed within it.
Precise air conditioning systems are often used in system-level cooling. They take hot air and dispose of it externally, circulating cooler air to reduce temperatures inside. Systems like these can be effective in data centers with consistent levels of demand and few HPC racks, but they are less likely to be used for more complex HPC applications.
Hydronic cooling systems can also be used to circulate cool air through data centers. These systems are inherently scalable, which has made them a popular choice amongst larger data centers, where demand is likely to grow over time. These are efficient systems and can handle higher temperatures than air conditioning systems. They’re also great for data centers that need to counteract variable heat on a day to day basis.
Another good option in system-level cooling is what’s known as free cooling. These systems use the temperature of the external environment to cool the inside of the building. They’re incredibly energy efficient, but of course they only work well when computational demand is relatively low, and they can only be used in locations where external temperatures are low enough to cool the data center sufficiently.
Flexibility is key in future-proofing of cooling technology
Data centers are changing dramatically right now, as they transform and adapt to the ever-increasing demands placed upon them.
Clients’ requirements are growing more and more complex, and as the capabilities of new AI applications starts to surpass what had been thought possible, so too do the extreme temperatures being generated in data center racks.
We’re not expecting a total overhaul of data center cooling technologies anytime soon, but we are predicting significant changes in the efficiency and power of this tech.
Alternative cooling technologies will likely become more commonplace, and as more and more companies look for energy-efficient ways of cooling that could reduce their carbon footprint, it’s likely that cooling tech will begin to improve in terms of efficiency.
Data centers could look to make their own efficiency improvements too. Those located near sources of cold water, or in colder climates, could start to capitalize on their location by making use of natural cooling methods available to them. Similarly, we could soon see higher levels of investment in data centers in such locations, as we prepare for the ongoing battle against the rising heat of data centers.
There have been suggestions that the temperatures at which data center equipment can safely run could be the next thing to change. Some equipment is now made to run at higher temperatures than the typical 70°F, and if equipment can be made to withstand higher temperatures without a risk of failure then this too could counteract the cooling problem that lies ahead.
It looks likely that traditional cooling methods will soon be a thing of the past, as demands on data centers grow and their need for reliant, highly efficient cooling increases. Sustainability remains of huge importance to data centers, and the prioritization of this will likely lead to significant changes in the way in which cooling is delivered.
It will continue to be vital that data centers are ready and able to adapt to new technology as and when it becomes available, allowing them to pass on the benefits of such tech to their clients.
What’s next for data center cooling?
It is now certain that further developments in artificial intelligence and the Internet of Things will lead the way in increasing data usage, and therefore our reliance on data centers. So, data centers can look ahead to a real upsurge in demand, and they (and their cooling systems!) need to be ready to manage that.
Datacenter cooling is already an important issue in the industry, but its prevalence will only increase as we continue to look for better, more energy efficient ways of keeping data centers at optimal temperatures.
The importance of data center cooling will only grow as we begin to depend more and more on data centers – no matter where we are in the world. So, keep your eyes peeled, because where there’s a need for innovation, it’s highly likely that some of the brightest minds of the industry will deliver over the coming years.
At TRG, we build the best datacenter experiences. We don’t accept anything below fault tolerance for our facilities, and we’ve made customer experience our top priority. Contact us to find out more about what we do.
