What Is a Data Center and How Does Its Cooling System Work?

Every time you open a banking app, stream a video, or send a message on social media, thousands of servers are working behind the scenes to process it. Those servers don't sit just anywhere. They are housed in a purpose-built facility called a data center. And one of the biggest challenges in running it is something that often goes unnoticed: heat.

This article explains what a data center is, why cooling is so critical, and how a data center cooling system works from end to end.

What Is a Data Center?

A data center is a physical facility that houses IT infrastructure, such as servers, storage devices, and networking equipment, along with the supporting systems that keep it running 24 hours a day, 7 days a week. Its purpose is to store, process, and distribute data and applications for businesses, governments, and public services.

A data center contains much more than servers. A facility typically consists of several core systems:

• IT systems: server racks, storage, switches, and routers.
• Electrical systems: utility power connections, generators, UPS (Uninterruptible Power Supply) units, distribution panels, and PDUs.
• Cooling systems (HVAC): equipment that removes heat and controls room temperature and humidity.
• Security and fire protection systems: access control, CCTV, smoke detectors, and gas-based fire suppression.
• Monitoring systems: BMS and DCIM platforms that track all operational parameters.

The reliability of a data center is commonly classified by Tier (Tier I to Tier IV) under the Uptime Institute standard. The higher the tier, the greater the system redundancy, including in the cooling system.

Why Do Data Centers Need Cooling?

Almost all of the electrical energy consumed by IT equipment ends up as heat. A conventional server rack can produce roughly 5 to 10 kW of heat, while racks built for artificial intelligence (AI) and high-performance computing can exceed 40 kW, and even reach 100 kW or more.

If this heat is not removed, equipment temperatures rise quickly. The consequences can include:

• Servers slowing down (thermal throttling) to protect themselves.
• Equipment shutting down automatically or suffering permanent damage.
• A shorter component lifespan.
• Service disruptions (downtime) that directly affect the business.

To prevent this, ASHRAE guidelines (through TC 9.9) recommend keeping the air temperature entering IT equipment at roughly 18 to 27 °C, with humidity controlled to avoid condensation and static electricity. In many data centers, cooling accounts for a very large share of energy use, often the second largest after the IT load itself.

The Basic Principle: Moving Heat, Not Creating Cold

Data center cooling is fundamentally a process of moving heat from IT equipment to the outside of the building. It happens in stages:

1. Cold air enters the front of the server and absorbs heat from the components.
2. The heated air exits from the back of the server.
3. The hot air is drawn back to the room's cooling units.
4. The heat is transferred to water or refrigerant.
5. The heat is finally rejected to the atmosphere through a chiller or cooling tower.

Every design variation in use today is simply a different way of carrying out these five steps as efficiently as possible.

Main Components of a Data Center Cooling System

1. CRAC and CRAH

A CRAC (Computer Room Air Conditioner) is a room cooling unit with its own compressor, working on a direct expansion refrigeration cycle. A CRAH (Computer Room Air Handler) has no compressor; it cools air using coils filled with chilled water supplied by a chiller. Both are known as precision cooling units because they can control temperature and humidity accurately.

2. Chiller

A chiller is a machine that produces chilled water, typically at around 7 to 12 °C in conventional systems, or at higher temperatures in modern, more energy-efficient designs. This chilled water is circulated to CRAH units or other cooling equipment, then returns warmer to be cooled again. Chillers can be air-cooled (rejecting heat directly to outdoor air) or water-cooled (rejecting heat through condenser water to a cooling tower).

3. Cooling Tower

In water-cooled systems, the cooling tower rejects heat from the condenser water to the atmosphere by evaporating a small portion of the water. This approach is far more energy-efficient than air-cooled systems, but it requires a water supply and good water treatment.

4. Pumps, Piping, and Valves

A network of pipes and pumps carries chilled water and condenser water throughout the facility. Redundant designs (such as N+1 or 2N configurations) for pumps and piping paths are essential so that the failure of a single component does not shut down cooling.

5. Raised Floor, Plenum, and Containment

In many designs, cold air is delivered through the space beneath a raised floor and emerges through perforated tiles in front of the racks. To keep cold and hot air from mixing, containment is used:

• Cold Aisle Containment (CAC): the cold aisle is enclosed so that cold air goes only into the servers.
• Hot Aisle Containment (HAC): the hot aisle is enclosed and hot air is channeled directly back to the cooling units.

Containment is among the most effective and least expensive ways to improve cooling efficiency, because it prevents air recirculation and bypass.

Cooling Approaches by Unit Location

Beyond the classic room layout, there are several other approaches depending on heat density:

• Room-based cooling: cooling units sit along the room's perimeter and cool the entire space. Suitable for low to medium density.
• Row-based (in-row) cooling: units are placed between racks, so the distance to the heat source is shorter and more controlled.
• Rack-based cooling: cooling happens at the rack level, for example with a rear door heat exchanger that absorbs heat right behind the servers.

Liquid Cooling: The Answer for High Density

Air has a limited ability to absorb heat. As rack density climbs, especially with AI servers packed with GPUs, air cooling alone is no longer enough. This is where liquid cooling comes in, because liquids absorb and carry heat far more effectively than air. The two most common approaches are:

• Direct-to-chip (cold plate): metal plates with liquid flowing through them are mounted directly on processors and GPUs to pull heat from the source.
• Immersion cooling: servers are submerged in a dielectric fluid that does not conduct electricity, so heat is absorbed directly by the liquid.

Many modern facilities use a hybrid approach: liquid cooling for the hottest components and air cooling for the rest.

Free Cooling and the Challenges of a Tropical Climate

In cooler climates, data centers can use cold outside air or water to cool the facility, allowing chillers to be partially or fully switched off. This technique is called free cooling, or economizer operation.

In Indonesia, the challenge is different. High temperatures and humidity year-round limit the opportunities for free cooling. Designs in tropical regions therefore typically emphasize:

• High-efficiency chillers and water-cooled systems.
• Tight containment to prevent air mixing.
• Raising the chilled water temperature and supply air temperature within safe limits.
• Careful humidity control to avoid condensation.
• Variable speed drives (VSDs) on pumps and fans so energy use follows the load.

Measuring Efficiency: PUE

The most widely used indicator of data center efficiency is PUE (Power Usage Effectiveness), the ratio of total facility energy to the energy used by IT equipment.

PUE = Total facility energy ÷ IT equipment energy

A PUE of 1.0 means all energy is used by IT equipment, with nothing spent on cooling or other supporting systems. The closer to 1.0, the more efficient the facility. Because cooling is the largest non-IT consumer, improvements to the cooling system almost always deliver the biggest reduction in PUE.

Redundancy and Reliability

Because downtime is extremely costly, cooling systems are designed with redundancy. Common configurations include N+1 (one spare unit beyond the required capacity) and 2N (two complete, independent systems). In high-tier data centers, even the piping paths, pumps, chillers, and power supplies to the cooling equipment are duplicated, so maintenance can be carried out without shutting down the facility (concurrently maintainable).

Conclusion

Data centers are the backbone of modern digital services, and their cooling systems are one of the main determinants of reliability and efficiency. The principle is simple: move heat from the servers to the outside of the building. But putting it into practice involves a set of integrated systems: CRAC/CRAH units, chillers, cooling towers, containment, and liquid cooling technology for AI workloads. As computing demand continues to grow, especially in tropical climates like Indonesia, expertise in mechanical HVAC for data centers will be increasingly valuable.

Frequently Asked Questions (FAQ)

What is the ideal temperature inside a data center? ASHRAE guidelines recommend an air temperature of roughly 18 to 27 °C entering IT equipment. Many operators now run toward the upper end of this range to save energy.

What is the difference between a CRAC and a CRAH? A CRAC uses a compressor and refrigerant inside the unit itself, while a CRAH uses chilled water from a chiller and has no compressor.

Does liquid cooling replace conventional air conditioning? Not entirely. In most facilities, liquid cooling is used together with air cooling in a hybrid system.

Why does PUE matter? PUE shows how much energy actually goes into computing versus how much is lost to supporting systems, making it a benchmark for both efficiency and operating cost.