Commissioning data center cooling systems

A data center's cooling system has one job that cannot fail: keep IT equipment within its temperature limits through every load condition, failure and maintenance event the facility was designed for. Commissioning is how owners find out, before production equipment goes live, whether the system will actually do that.

Data center commissioning is more rigorous than typical building commissioning. It tests systems under real heat loads, deliberately fails equipment, and verifies that mechanical, electrical and controls systems work together. This guide outlines the process and the tests that matter most.

The commissioning framework

Data center commissioning follows the general process described in ASHRAE Guideline 0 and ASHRAE Standard 202, adapted for mission-critical facilities. It begins at the start of design, not at the end of construction. See the commissioning process from OPR to turnover for the general framework.

Key documents include:

• Owner's project requirements (OPR): availability targets, redundancy, temperature and humidity envelope, efficiency goals, operational expectations
• Basis of design (BOD): how the design meets the OPR
• Commissioning plan: scope, roles, schedule, test levels and procedures
• Test scripts: step-by-step procedures with acceptance criteria
• Issues log: deficiencies and their resolution

The five levels

The data center industry commonly organizes commissioning into five levels, each building on the previous one. Terminology varies among owners and commissioning providers, but the sequence is widely used.

Level 1: factory acceptance

Factory tests verify capacity, efficiency, controls and alarms before equipment ships. Chillers may be tested on a factory test stand at specified conditions. CDUs and modular systems can be tested as complete assemblies. Problems found here are much cheaper to fix than on site.

Level 2: installation verification

Equipment is inspected on arrival and after installation. Checklists confirm model numbers, clearances, piping, electrical connections, sensor locations and labelling.

Level 3: startup and pre-functional

This is where systems first run. Activities include:

1. Pressure testing and flushing of chilled water and liquid cooling loops
2. Water treatment and fluid fill
3. Manufacturer startup of chillers, pumps, towers, CRAHs and CDUs
4. Controls point-to-point verification, confirming every sensor reads correctly and every output moves the right device
5. Initial water balancing and air balancing

Level 4: functional performance testing

Each system is tested through every mode in its sequence of operation. For cooling this includes:

• Normal operation across the load range
• Economizer, partial economizer and mechanical modes, and transitions between them
• Failure of each chiller, pump, tower fan, CRAH and CDU in turn, with standby response
• Sensor failure responses
• Alarm generation and communication to BAS and DCIM
• Restart after loss of power to individual equipment

See functional performance testing.

Level 5: integrated systems testing

Integrated systems testing (IST) brings everything together. With heat load applied to the data hall, the team simulates utility power failure and observes the full chain: UPS carrying IT and critical mechanical loads, generators starting, transfer switches operating, chillers restarting, thermal storage carrying the gap, and the system returning to normal. Recovery when utility power returns is also tested.

In Canada, integrated testing of fire protection and life safety systems follows CAN/ULC-S1001, which is distinct from but often coordinated with the data center IST. See integrated systems testing and CAN/ULC-S1001.

Load testing

Cooling cannot be meaningfully tested without heat. Before IT equipment arrives, temporary heat loads are installed:

• Rack-mounted load banks that mimic server airflow and heat
• Floor-standing heater fans or load banks in rows
• Liquid heat load simulators for direct-to-chip and CDU systems

Load banks should be distributed to represent the design layout, not clustered in one area. See load bank testing for cooling validation.

What to measure during testing

• Rack inlet temperatures across the data hall, at multiple heights
• Supply and return temperatures at every cooling unit
• Chilled water and condenser water temperatures and flows
• Chiller, pump and fan power
• Room and plenum pressures
• Time to recover after failure events
• Peak temperatures reached during transients

Data loggers with fast sampling are needed to capture transient events such as the temperature rise during a chiller restart.

Common gaps

1. Commissioning starts too late, so design issues are found during construction.
2. Testing at low load only, missing capacity or control problems at design load.
3. Failure tests skipped because of schedule pressure.
4. Controls not fully verified, especially alarms and DCIM integration.
5. Seasonal modes untested, such as full economizer or peak summer operation. Plan seasonal testing.
6. Poor documentation of results, leaving operators without a baseline.

Turnover and training

At the end of commissioning, operators need complete documentation: as-built drawings, sequences of operation, test results, trend baselines, O&M manuals and a systems manual. Hands-on training covering normal operation, failure response and emergency procedures should be delivered by the people who know the systems.

Frequently asked questions

Who should perform data center commissioning?

An independent commissioning authority working for the owner is best practice, with contractors and manufacturers executing tests under their direction. Independence avoids conflicts of interest.

How long does it take?

That depends on facility size and complexity. Level 4 and Level 5 testing alone can take weeks for a large facility. Schedule commissioning time realistically from the start, rather than compressing it at the end.

Key takeaways

• Data center commissioning starts at design with a clear OPR and continues through turnover.
• The five levels run from factory tests to integrated systems testing under load.
• Failure and transient tests under realistic heat load are the heart of cooling commissioning.
• Load banks must be distributed to represent the real layout.
• Thorough documentation and training give operators the baseline they need.