Data center cooling controls and DCIM integration

The cooling equipment in a data center is only as good as the controls that run it. Well-designed controls keep rack inlets within limits, use free cooling whenever possible, respond to failures automatically and give operators a clear picture of what is happening. Poor controls cause units to fight, waste energy, mask failures and generate floods of meaningless alarms.

Data centers usually have several overlapping monitoring and control systems: a building automation system (BAS), an electrical power monitoring system (EPMS), equipment-level controllers and a data center infrastructure management (DCIM) platform. This guide explains how cooling controls should work and how these systems fit together.

Control strategies that matter

Control on rack inlet temperature

Older systems controlled each cooling unit on its own return air temperature. That approach keeps the return air at setpoint but says nothing about what servers actually receive. Modern practice controls supply air temperature at the cooling unit and uses rack inlet temperature sensors to confirm conditions, or to reset setpoints. The ASHRAE TC 9.9 recommended range of 18 to 27 °C (64 to 81 °F) applies at server inlets, so that is where it should be measured.

Separate airflow and temperature control

Airflow and temperature are two different jobs:

• Fan speed should track IT airflow demand, often by maintaining a small positive differential pressure between cold aisle and hot aisle in contained rooms, or a set underfloor pressure in raised-floor rooms.
• Cooling coil valves or compressors should control supply air temperature.

Combining both into a single return temperature loop causes poor performance.

Coordinated control across units

Distributed cooling units should operate as a team. Strategies include a shared supply temperature setpoint, fan speeds controlled together from common pressure measurements, and humidity control centralized rather than on every unit. This prevents one unit from humidifying while its neighbour dehumidifies, or one unit running at full speed while the next idles.

Plant optimization

Chiller plants, economizers and pumps should be sequenced to minimize total energy while meeting load, with chilled water reset based on actual demand. See chiller plant sequencing and optimization.

Liquid cooling controls

CDUs control secondary supply temperature, flow and pressure, often with dew point protection that raises supply temperature if room humidity climbs. See coolant distribution units.

Sensors

Sensor accuracy and calibration matter: a drifting sensor on a control loop can waste energy or cause overheating without anyone noticing. See sensor selection, location and calibration.

The systems involved

• Equipment controllers: built into chillers, CRAHs, CDUs and in-row coolers; run local control loops.
• BAS: supervises and coordinates mechanical equipment, runs plant sequences and generates alarms. Typically uses BACnet.
• EPMS: monitors electrical distribution, UPS, generators and switchgear. Often uses Modbus or IEC 61850.
• DCIM: combines facility data with IT asset information (rack locations, power per rack, capacity) to support capacity planning, change management and reporting.

The BAS should remain the system that controls mechanical equipment. DCIM is usually a monitoring and planning layer that reads data from the BAS, EPMS and IT systems; giving DCIM direct control of cooling equipment adds risk unless carefully engineered.

Communication protocols

See BACnet basics for building owners.

Integration steps

1. Define requirements: what data each system needs, who controls what, required update rates and alarm responsibilities.
2. Create a points list: every point to be shared, with names, units, protocol addresses and polling intervals.
3. Agree naming conventions so that a CRAH or rack has the same identifier in the BAS, DCIM and asset records.
4. Design the network: segregated networks for building systems, firewalls between OT and IT, and controlled remote access.
5. Configure and test: verify every point end to end, checking values, units and alarm states.
6. Document: network diagrams, points lists, integration settings and ownership.

Alarms

Data centers can produce thousands of alarms a day if not managed. Good alarm management:

• Prioritizes alarms by consequence (critical, urgent, advisory)
• Suppresses consequential alarms when a root cause alarm is active
• Routes alarms to the right people with escalation
• Is reviewed regularly to remove nuisance alarms

See alarm management in building automation.

Cybersecurity

Connecting cooling controls to IT networks and DCIM increases exposure. A compromised BAS in a data center could affect cooling and therefore IT availability. Apply network segmentation, strong authentication, patching, secure remote access and monitoring. Standards such as the ISA/IEC 62443 series provide a framework.

Verifying controls

Controls should be verified during commissioning through point-to-point checks, sequence testing, failure testing and alarm verification, and then checked periodically through trend reviews. Control problems, such as disabled economizers, units in manual override and drifting sensors, are among the most common findings in existing data centers.

Frequently asked questions

Should DCIM control the cooling?

Generally, DCIM should monitor and inform, while the BAS and equipment controllers control. Some advanced systems use DCIM data, such as rack inlet temperatures, to reset BAS setpoints, but control authority and failsafe behaviour must be clearly defined.

How many rack inlet sensors do I need?

Common practice is several sensors per rack at different heights on representative racks, with at least one per rack in high-density areas. The goal is to catch hot spots without creating an unmanageable data volume.

Key takeaways

• Control airflow and temperature separately, and confirm conditions at rack inlets.
• Coordinate cooling units as a team and centralize humidity control.
• Keep mechanical control in the BAS; use DCIM for monitoring, planning and reporting.
• Plan integration with a points list, consistent naming and secure network design.
• Manage alarms carefully and verify controls during commissioning and in operation.