When a data center starts cooling chips with liquid, something has to sit between the messy building water system and the pristine coolant that flows through the cold plates. That something is the coolant distribution unit. It is the hydraulic gatekeeper of liquid cooling, isolating the two loops, moving heat between them, and precisely controlling the temperature, flow, and pressure of the coolant the servers actually see. This guide explains what a CDU does, its pumps and heat exchanger, and the setpoints a SCADA system regulates and alarms.
Coolant Distribution Unit (CDU) in one line: A coolant distribution unit, or CDU, is the hydraulic interface between a facility's water system and the clean coolant loop that feeds cold plates or other liquid-cooled IT equipment. It uses a heat exchanger to transfer heat from the technology cooling loop into the facility loop while keeping the two fluids separate, and its pumps circulate the clean coolant to the racks at a controlled temperature, flow, and pressure. In effect, the CDU protects and regulates the delicate loop that touches the servers.
Liquid cooling actually involves two separate water circuits, and the CDU is what joins them without mixing them. On one side is the facility water loop, the building's chilled or process water, which may be relatively cool but is not clean enough or controlled enough to send through the fine channels of a cold plate. On the other side is the technology cooling system loop, sometimes called the secondary loop, which carries carefully treated, filtered coolant through the cold plates on the processors. If the dirty facility water ever reached the cold plates, sediment and corrosion products would clog and damage them, so the two must never mix.
The CDU keeps them apart with a heat exchanger, a device with two sets of passages so close together that heat passes readily from one fluid to the other while the fluids themselves stay completely separate. Warm coolant returning from the racks gives up its heat through the exchanger to the cooler facility water, then returns to the racks cooled down. The facility water, now warmer, carries the heat away to the chillers or heat-rejection plant. This isolation is the CDU's core purpose: the servers see only the clean, controlled secondary loop, while the building's water does the heavy lifting of ultimately rejecting the heat.
This division of labour also decouples the two sides so each can be optimised. The secondary loop can be run at exactly the temperature and pressure the IT hardware wants, independent of whatever the facility water is doing, and the facility side can be managed for efficiency without worrying about the servers. CDUs come in different sizes, from rack-mounted units that serve a single rack or row up to large floor-standing units that feed many racks, but the principle is the same at every scale: one clean loop for the hardware, one facility loop for heat rejection, and a heat exchanger between them.
Inside the CDU, the pumps are what actually push the clean coolant around the secondary loop to the cold plates and back. Because losing circulation to hundred-kilowatt racks would be catastrophic, CDUs typically carry redundant pumps so that if one fails the other keeps the loop moving, and the pumps are often variable-speed so the unit can raise or lower flow to match the load. The pumps must deliver enough flow and pressure to reach every cold plate in the loop while staying within the limits the hardware and plumbing can tolerate, which is why flow and pressure control is central to how a CDU operates.
The heat exchanger is the heart of the unit and sets how much heat the CDU can move. Its capacity, together with the temperature of the facility water available, determines how cold the secondary loop can be held for a given heat load. On the facility side, a control valve usually modulates how much facility water passes through the exchanger, which is the CDU's main lever for holding the secondary supply temperature steady as the IT load rises and falls. When the racks work harder and the return coolant gets warmer, the CDU opens the valve to pull more facility water and hold the setpoint.
The CDU also protects the clean loop it serves. Filtration removes particulates that would otherwise accumulate in the narrow cold-plate channels, and the unit manages the coolant's condition and inventory, including make-up and expansion. Many CDUs run the secondary loop above the room dew point to prevent condensation on cold surfaces near the racks. All of this housekeeping is what lets the same clean fluid circulate safely through delicate hardware for a long time, and it is why the CDU, not the cold plate, is the component operators watch most closely.
A CDU is a control device, and its behaviour is defined by a handful of setpoints that a monitoring or SCADA system watches and, in many deployments, helps regulate. The most important is the secondary supply temperature, the temperature at which coolant is delivered to the cold plates. It is held to a target that is cool enough to keep the chips safe but warm enough to avoid condensation and to run the facility efficiently, and the CDU modulates its facility-side valve to maintain it. Deviation from this setpoint is a first-order alarm, because a supply temperature drifting up means the hardware is losing its cooling margin.
Flow and pressure are the next tier of setpoints. The CDU regulates coolant flow, often to hold a target differential pressure across the loop so every cold plate gets adequate flow, and it alarms if flow or pressure falls outside the safe band. Pump status and speed are monitored so that a pump failure and the failover to its backup are visible immediately. On the facility side, the supply and return temperatures and the valve position tell operators whether the building has enough cooling capacity to feed the CDU, and a facility loop that is too warm will show up as the CDU struggling to hold its secondary setpoint.
These are precisely the signals a cloud SCADA platform such as Merobix is designed to handle. A CDU exposes its temperatures, flows, pressures, pump states, valve positions, and leak or level alarms, usually over an industrial protocol, and a SCADA system ingests them as tags, trends them, and raises alerts when a setpoint is missed. Operators, whether standing in the facility or overseeing a fleet of sites remotely, then see each CDU in the same dashboards and alarm workflows as the pumps, tanks, and process loops of any other monitored plant. Because the CDU is the single point that keeps an entire liquid-cooled row alive, having it continuously monitored and remotely visible is what turns liquid cooling from a fragile novelty into managed infrastructure.
A CDU sits between the facility water system and the clean coolant loop that feeds cold plates, using a heat exchanger to move heat between them while keeping the two fluids separate. Its pumps circulate the clean secondary coolant to the racks at a controlled temperature, flow, and pressure. It protects the delicate loop the servers rely on and regulates the conditions that keep the hardware safe.
The facility water is not clean or controlled enough to send through the fine channels of a cold plate; its sediment and corrosion products would clog and damage them. The technology cooling loop uses carefully treated, filtered coolant just for the hardware. The CDU's heat exchanger lets heat pass between the loops without the fluids ever mixing, so the servers only ever see clean, controlled coolant.
The primary setpoint is the secondary supply temperature delivered to the cold plates, held cool enough to protect the chips but above the dew point to avoid condensation. The CDU also regulates coolant flow and loop differential pressure so every cold plate gets adequate flow. A SCADA system watches these setpoints along with pump status and alarms when any drifts out of its safe range.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.