In-row and rear-door heat exchanger cooling

Traditional data center cooling pushes cold air from perimeter units across a room and hopes it reaches the server inlets before it mixes with hot exhaust. That works at modest densities, but as rack loads climb the distance between the heat source and the cooling coil becomes the problem. Close-coupled cooling shortens that distance by putting the coil right beside or right behind the rack.

The two most common close-coupled approaches are in-row coolers and rear-door heat exchangers (RDHx). Both use chilled water or refrigerant to absorb heat within a metre or so of the IT equipment. They solve different problems, and choosing between them depends on rack density, room layout, water availability and how much the owner wants to disturb an existing white space.

Why close-coupled cooling exists

Room-level cooling moves large volumes of air over long paths. Along the way, cold supply air bypasses the servers and hot exhaust recirculates to the inlets. Operators compensate by lowering supply temperatures and oversupplying air, which wastes fan and chiller energy. Containment fixes much of this, but at higher densities the airflow a perimeter unit must deliver, and the static pressure under a raised floor, become limiting.

Close-coupled systems capture heat at the source. The air path is short, so fan energy drops, mixing is reduced and the coil sees warmer return air, which improves coil effectiveness and allows warmer chilled water.

In-row coolers

An in-row cooler is a cabinet-sized unit, typically 300 mm (12 in) or 600 mm (24 in) wide, placed in the row between IT racks. It draws hot air from the hot aisle, passes it through a coil, and discharges cooled air into the cold aisle. Variable-speed EC fans adjust airflow to match load, usually controlled on rack inlet temperature or on differential pressure across a contained aisle.

Where in-row units work well

• Slab-floor rooms with no raised floor plenum
• Rows of 10 to 30 kW racks, especially with hot aisle containment
• Mixed-density rooms where a few rows need more capacity than the perimeter system can deliver
• Pods that need to be deployed in stages as load grows

Limitations

In-row coolers take up rack positions, which reduces usable white space. They also bring water or refrigerant piping into the row, which requires leak detection, drip containment where piping runs overhead, and careful routing. Because each unit serves only nearby racks, failure of one unit has a local effect, so redundancy is usually designed per row or per pod (for example N+1 within a contained aisle).

Rear-door heat exchangers

A rear-door heat exchanger replaces the back door of a rack with a finned coil. Server exhaust passes directly through the coil and leaves the rack at or near room temperature. In effect, the rack becomes room-neutral: the room sees little of the heat the servers produce.

Passive and active doors

Capacities depend heavily on entering water temperature, server airflow and the temperature rise across the servers. Always check manufacturer data at your actual operating conditions rather than catalogue headline ratings.

Where RDHx works well

• Retrofits in existing rooms where only some racks are increasing in density
• Rooms where adding containment is impractical
• Sites wanting to raise room temperature or eliminate perimeter units entirely
• As a supplement to direct-to-chip liquid cooling, to capture the residual air-cooled heat from memory, drives and power supplies

Comparing the options

Water temperature and the fluid loop

Both technologies work best with elevated chilled water temperatures. Because the coil sees hot exhaust air directly, entering water in the range of roughly 14 to 20 °C (57 to 68 °F) is common, which keeps the coil above the room dew point and avoids condensation. Designing above dew point matters: these units are intended for sensible cooling only, and most are not provided with condensate drainage.

Many designs isolate the rack-level loop from the building chilled water with a coolant distribution unit. The CDU controls supply temperature above dew point, limits the volume of water that could leak into the white space, and allows a cleaner, filtered secondary loop. Refrigerant-based doors and in-row units (pumped refrigerant) are an alternative where owners want no water in the room; these must be designed to CSA B52 refrigerant quantity limits for the occupied space.

Warmer water also extends economizer hours. In most Canadian climates a plant serving close-coupled cooling at 18 °C (64 °F) supply can run on free cooling for a large share of the year.

Design checklist

1. Establish actual rack loads and airflow, not nameplate. Ask for measured or vendor-supplied IT power and server airflow per kW.
2. Confirm the operating envelope you intend to hold, normally the ASHRAE TC 9.9 recommended range of 18 to 27 °C (64 to 81 °F) at server inlets.
3. Select units at the real entering water temperature and flow, and at the server temperature rise expected.
4. Set the design water temperature above the maximum expected room dew point, with margin.
5. Plan redundancy: N+1 per aisle for in-row, and a strategy for what happens to a rack if its door loses water.
6. Lay out piping with isolation valves at each unit, flexible hoses rated for the pressure, drip trays where needed and leak detection cable under or beside piping.
7. Coordinate controls: fan speed, valve position and alarms into the BAS or DCIM.
8. Plan commissioning with heat load testing before IT equipment arrives.

Commissioning and verification

Close-coupled systems should be tested under load. Load banks or heater fans placed in the racks allow the team to verify capacity, control response and failure behaviour, such as what happens to inlet temperatures when one in-row unit stops. Water flow to each unit should be balanced and recorded, and control loops tuned so that fans and valves do not hunt. See load bank testing for cooling validation for how these tests are set up.

Frequently asked questions

Can I add rear doors to an existing room without shutting it down?

Often yes. Doors can usually be installed rack by rack, with water connections made from overhead or underfloor piping that is installed and pressure-tested in advance. The main constraints are rack compatibility and getting piping to each position safely.

Do in-row coolers need hot aisle containment?

They will run without it, but performance drops because cold supply air mixes with hot exhaust before reaching the coil. Containment lets the unit see high return temperatures and keeps capacity and efficiency close to rated values.

Is water at the rack a serious risk?

It is a manageable risk. Good practice is a low-volume secondary loop, quality hose connections, leak detection with automatic isolation where justified, and keeping supply water above dew point so condensation is not an issue.

Key takeaways

• In-row coolers and rear-door heat exchangers cool at the rack, cutting fan energy and reducing mixing.
• In-row units suit contained rows and slab floors; rear doors suit retrofits and rack-by-rack density increases.
• Both are sensible-only devices and should be supplied with water above the room dew point.
• A CDU is often used to isolate and control the rack-level loop.
• Size units at real operating conditions, plan redundancy deliberately and verify performance with load testing before go-live.