Data Centre Cooling Guide: Selecting the Right System for Your Server Room or Data Centre

Energy-efficient and effective cooling is one of the most operationally critical decisions in any server room or data centre project. More so with the rise of AI server deployment and the increasing kW loads this can bring to any server rack installation.

Any electronic device generates heat and, within a server room or data centre, this can include not just the IT loads (servers, networking devices, power cables and distribution) but also uninterruptible power supplies. Without a correctly specified and installed cooling system, ambient temperatures can rise rapidly, and hotspots can be more prevalent in areas of the setup with less airflow (behind server cabinets). As temperatures rise, so does thermal stress on equipment; more energy is required for cooling, and the risk of unplanned downtime and system failures increases.

The challenge is not simply to install a cooling system that works from day one, but one that is sized to support future changes in the IT load configurations. Resilience is also important, with the preferred solution being an N+1 arrangement to allow for routine maintenance and ensure uptime.

Server Room Environments supplies and installs a complete range of cooling solutions for facilities of all sizes, from small comms rooms and edge deployments through to multi-room data centres. Our systems include traditional air conditioners, in-row systems, and liquid-cooling solutions

In this guide, we outline the principal cooling technologies available, how to compare them, and the key factors that should inform any cooling project’s initial design or system upgrade.

Why Cooling Matters in Server Rooms and Data Centres

IT hardware, for example, traditional and AI servers, must operate within the defined thermal tolerances specified in their datasheets.

ASHRAE Class A1 guidelines recommend server inlet temperatures between 18°C and 27°C, with many operators targeting 20–24°C to provide additional headroom. It is important to note that this is the cooled air to be supplied to the front of server cabinets with an expected Delta-T gain of at least 10°C as air passes through the server cabinet to the rear exhaust area.

When ambient temperatures exceed these levels, performance degrades before system failure occurs, often presenting as intermittent errors, increased power draw and shortened component lifespans rather than immediate outright failure.

For more information see: https://xp20.ashrae.org/datacom1_4th/ReferenceCard.pdf

Cooling also directly influences Power Usage Effectiveness (PUE), the ratio of total facility energy to IT equipment energy. A poorly designed cooling strategy forces cooling systems to work harder than necessary, increasing energy usage costs without improving outcomes.

Conversely, a well-designed cooling strategy with appropriate (hot and cold aisle) containment and controls can achieve PUE values below 1.5 (industry average), and with direct free cooling and immersion systems able to reach 1.1 or lower.

For more information see: https://www.statista.com/statistics/1229367/data-center-average-annual-pue-worldwide/?srsltid=AfmBOopW0UM-S1-iQdX1kV0XiYF0lVHnWpQ7WKwqmzIenyU6tRQBzz_u

For data centre operators managing colocation spaces, providing cloud services or regulated public sector, healthcare or financial industry IT environments, cooling performance can also be a major factor within Service Level Agreement (SLA) compliance, in addition to uptime commitments and audit readiness.

Cooling Methods: An Overview

No single cooling technology suits every facility. The right approach depends on server rack density, room size and layout, available infrastructure, required resilience level and projected growth. The table below summarises the main options:

Precision Air Conditioning and CRAC Units

Computer Room Air Conditioners (CRACs) as well as precision air conditioning (PAC) solutions have been the standard approach for computer and server room cooling for decades. Unlike standard HVAC (Heating, Ventilation, and Air Conditioning) systems, precision units are designed specifically for IT environments, supplying 24/7, accurate temperature and humidity control and providing the stable operating environment that vulnerable IT server equipment requires.

CRAC units draw room air across a direct expansion (DX) refrigeration coil and return conditioned air at a controlled temperature. Precision air conditioning systems operate on the same principle but provide increased flexibility in installation configuration, redundancy and integration with building management systems (BMS).

CRACs and PAC solutions are often referred to as ‘Splits’ as they consist of one or more header units in the room (often wall-mounted or floor-standing) and an external heat exchanger. From a room and server rack layout point of view, the preferred installation method is to place the header unit at the rear and top of the server racks; this then pushes cold air over the racks to the front of the servers. This cool air is drawn into the server racks and exhausted at the rear of the cabinets, where it is drawn into the header unit for cooling. Header units are typically installed at the front of the server racks, with no way to direct the hot exhaust air back in a return path for cooling. For efficient cooling, server racks should also have their side panels and doors installed and closed. Blanking panels should be installed where there is unused ‘U’ capacity in the server racks. This prevents cool and hot air from blending within the cabinets and room space.

For smaller facilities, including comms rooms and single-room server environments, Daikin or Mitsubishi wall-mounted or floor-standing systems are the most commonly used. Sometimes overhead ceiling-suspended precision units may be installed, but these should be avoided where possible. Wall-mounted units should also be installed with drip trays to catch any water leakage from the header unit pumps, should a problem occur. This can also be monitored remotely using a suitable environmental monitoring system and a spot or water-leakage detection rope. Key specification considerations for all precision cooling installations are N+1 redundancy to eliminate single points of failure, correct unit sizing to prevent short-cycling, and appropriate placement to support effective airflow management throughout the room.

For later server rooms and data centres, an APC by Schneider or Airsys precision cooling system is our preferred solution. The Airsys CYBERMOD series combines precise temperature and humidity control with modular design technology, simplifying system expansion. New units can be integrated into an existing CYBERMOD arrangement as cooling requirements grow, with Energy Efficiency Ratios (EERs) exceeding 20 across the range. Cooling capacities range from 44kW up to 215kW in chilled water (CW) configuration.

In-Row Cooling and Close-Coupled Solutions

As rack densities increase beyond 10kW per rack, room-level cooling becomes less effective. The distance between a wall-mounted or floor-mounted cooling unit and the heat source grows, allowing hot exhaust air from high-density racks to mix with conditioned supply air before it reaches equipment inlets. The result is reduced cooling efficiency and localised hot spots where temperatures can exceed acceptable ambient thresholds, even when the overall room temperature appears within range.

In-row cooling systems tackle this approach head-on by positioning cooling units within the server row, perpendicular to the racks. Cool air is delivered at the point of need, and hot exhaust air is captured before it can escape into the wider room environment and is routed to an external heat exchanger. This approach dramatically improves cooling effectiveness at high densities and scales incrementally as additional racks are added.

Close-coupled solutions, including rear-door heat exchangers and overhead cooling units, operate on the same principle: removing heat as close to the source as possible. Both approaches integrate effectively with hot-aisle and cold-aisle containment solutions, further improving efficiency by separating supply and return airflow.

Air Handling Units (CRAHs) and Chilled Water Systems

For larger data centres with higher total heat loads, Computer Room Air Handlers (CRAHs) supplied with chilled water from a central chiller plant are a well-established solution. Unlike DX precision units, which generate cooling independently, CRAHs rely on a chilled water supply, typically from roof-mounted or external chillers, to cool air before it is distributed throughout the data hall.

Chilled water systems provide significantly higher capacity advantages and can be configured with a high degree of redundancy at both the chiller and air handler levels. They also support integration with free-cooling economisers, which can substantially reduce mechanical refrigeration hours and improve PUE during periods when outdoor temperatures allow it.

Server Room Environments supplies and installs CRAHs and chiller systems as part of complete data centre cooling designs, with nationwide engineering support for both installation and ongoing planned preventative maintenance.

Direct Free Cooling

For installations in the UK where outdoor temperatures are frequently lower than the internal data centre temperatures, direct free cooling (DFC) can be introduced to supplement or replace mechanical refrigeration for a significant portion of the year. DFC introduces filtered external air directly into the server room or data centre when ambient conditions allow, reducing or eliminating the need to run compressor-based cooling during those periods.

The energy savings from DFC can be substantial. In the UK climate, a well-designed DFC system can provide economiser-mode cooling for a significant proportion of annual operating hours, with corresponding reductions in energy consumption and PUE. Airsys DFC-capable systems supplied by Server Room Environments combine DX mechanical cooling with direct free-cooling, automatically switching between modes depending on ambient conditions.

Hot Aisle and Cold Aisle Containment

Aisle containment is one of the most cost-effective improvements available to any existing data centre or server room. Without containment, hot exhaust air from the rear of server racks mixes with cold supply air before reaching equipment inlets, reducing the effective temperature differential and forcing cooling systems to work harder to maintain setpoints.

The principle of containment is fairly straightforward. Racks are arranged so that equipment exhausts hot air into a designated hot aisle, while cold supply air is delivered into a separate cold aisle facing equipment inlets. By physically separating supply and return airflows using blanking panels, containment curtains, solid doors or full enclosures, the mixing of hot and cold air is prevented.

Cold-aisle containment captures and directs cold supply air to equipment inlets. Hot-aisle containment captures exhaust air and channels it directly back to the cooling unit returns. Full enclosures provide the highest level of separation, especially effective within high-density computing settings in which even small amounts of air mixing represent a meaningful efficiency loss.

Facilities with proper hot/cold aisle containment typically achieve measurable reductions in cooling energy consumption. Substantial PUE improvements of 0.2 to 0.5 are achievable in many cases, which translate directly into reduced operating costs and improved sustainability metrics. Server Room Environments works with a UK partner to manufacture bespoke containment solutions that accommodate non-standard room layouts, mixed rack heights, and complex airflow configurations.

Environmental Monitoring for Cooling Performance

A cooling system is only as reliable as the visibility operators have into its performance. Without 24/7 environmental monitoring, thermal issues (temperature and humidity) can develop undetected, and by the time hardware errors or equipment failures surface, damage may already have occurred.

Temperature is the most fundamental metric. Monitoring should cover both cold-aisle supply and hot-aisle return temperatures, as the differential between them reflects how effectively heat is being removed from equipment.

Humidity monitoring is equally important and frequently underestimated. Below approximately 40% relative humidity, the risk of electrostatic discharge (ESD) increases. Above 60%, condensation becomes a concern, particularly around cooling coils and cable management. Precision cooling systems manage humidity alongside temperature, but independent monitoring confirms that control is being maintained under dynamic load conditions, especially within an AI server environment.

Air quality monitoring, including particulate levels, is also relevant for facilities targeting ISO 14644-1 Class 8 cleanliness. Dust accumulation on heatsinks and internal components acts as insulation, reducing heat transfer efficiency and forcing cooling systems to work harder than necessary.

Server Room Environments supplies and integrates environmental monitoring solutions, including temperature, humidity, air quality, water leakage and power monitoring as part of complete data centre cooling and infrastructure projects. These systems provide 24/7 automated alerts via email, SMS and integration with BMS and DCIM platforms.

Selecting the Right Cooling Strategy

Cooling specification should begin with an accurate kW heat load assessment of the entire facility. Total installed IT power, average utilisation, peak demand and projected growth over a 2-5 year horizon should all feed into the cooling capacity calculation. Undersizing creates operational risks and potentially downtime; oversizing wastes capital and inflates running costs without improving outcomes.

Redundancy requirements should reflect the criticality of the systems being protected. For IT environments in which downtime can lead to direct financial or operational consequences, N+1 cooling redundancy should be considered. In this configuration, the failure of a single header unit should not compromise the facility. The installation of two or more systems supports concurrent maintenance, a requirement if the server room or data centre is operating at a specific Tier Level (Uptime Institute). For mission-critical environments or those operating under strict uptime SLAs, 2N cooling architectures provide additional resilience.

Scalability is a consideration that is sometimes missed at the design stage. Whilst on-site server rooms and some data centres may operate a hybrid mix of on-site and cloud-based services, power demands are now rising as AI server technologies are adopted. Many companies are deploying their own AI servers because of worries about data sovereignty and the need to protect their intellectual property.

A server room or data centre housing 10-15 racks today may need to accommodate 20-30 server cabinets within, say, 3-5 years, and cooling infrastructure that requires a complete rip-and-replace to add capacity represents both an operational and financial risk. Modular precision cooling systems, in-row solutions, and phased chiller plant designs offer increased flexibility for incremental expansion over the facility’s required working life.

The full lifecycle cost should inform specification decisions rather than upfront capital outlay. Energy-efficient cooling systems, intelligent controls, hot-aisle, cold-aisle containment, and environmental monitoring should consistently deliver strong returns over a five-year period.

Summary

Server Room Environments supplies a complete range of cooling solutions for comms rooms, server rooms and data centres of all sizes. This includes precision air conditioning, in-row cooling, CRAHs, chillers, direct free-cooling systems, hot-aisle and cold-aisle containment, and integrated environmental monitoring, all designed, installed, and maintained by manufacturer-certified engineers based across the UK.

Cooling projects range from single-unit replacements in existing server rooms to full data centre design and build, with free site surveys to assess IT kW and heat loads, airflow, redundancy, and efficiency requirements. Our clients typically achieve reductions in total cost of ownership of 50% or more when upgrading legacy cooling systems to current-generation solutions.

For more information or advice on server room and data centre cooling solutions, or to arrange a free site survey, please contact our Projects Team.