Plan data-centre cooling from the equipment inlet

List the server, storage and network equipment classes expected in each area and obtain the current manufacturer environmental requirements. Define the normal control range, alert range and shutdown or protection limits at the correct measurement point. ASHRAE data-centre guidance uses equipment inlet conditions as the common reference and connects those conditions to measurement, placement, airflow and reported heat load.

Sources: ASHRAE

Do not turn an allowable limit into the normal target without an engineering decision. Allowed excursions, reliability expectations, warranty terms, humidity control, contamination and the mix of equipment all matter. Record which requirement governs when equipment from several manufacturers shares a room. Keep temporary construction or commissioning equipment out of the final control assumptions unless it will remain in normal service.

Sources: ASHRAE

Convert the IT deployment plan into heat at the rack and row level. Record current load, expected peak, growth stage, equipment diversity and the period each stage will operate before the next expansion. Average room density can hide a small number of high-density racks that control the design. Name the source of each load figure and distinguish measured input, manufacturer data and planning allowance.

Sources: ASHRAE, Natural Resources Canada

Check the thermal consequence of partial loading. Fans, pumps, chillers, heat exchangers and containment can behave differently at low load than at design load. A new hall may spend months below its final density. The control sequence should keep stable inlet conditions without excessive bypass airflow, short cycling or unnecessary pumping while leaving a clear path to later deployment stages.

Sources: Natural Resources Canada, ASHRAE

Trace supply air from the cooling unit to the equipment inlet and return air back to heat rejection. Look for open rack spaces, cable openings, missing panels, poor containment, blocked tiles, recirculation and short circuits. Measure at representative rack inlets instead of relying on one room sensor. Correcting the path can recover useful capacity and improve temperature consistency before another cooling unit is installed.

Sources: ASHRAE, Natural Resources Canada

Document how doors, containment, smoke controls and maintenance work affect airflow. A normal diagram may not represent the room while a row is open, a fan is isolated or a filter is being changed. Define the permitted work state, temporary monitoring and restoration check. Keep the arrangement consistent with fire-protection and life-safety requirements established by the responsible designers and authorities.

Sources: ASHRAE

For direct-to-chip or other liquid-cooled equipment, record fluid requirements, supply and return conditions, heat load, pressure, flow, water quality, materials compatibility, filtration, leak detection, isolation and service access. Define the boundary between the facility loop, coolant distribution unit and IT equipment. Confirm which party owns each sensor, alarm, maintenance task and warranty condition.

Sources: ASHRAE

Plan how the system behaves during loss of pumps, loss of heat rejection, utility transfer, generator start, control failure and a leaking branch. State the thermal ride-through available at the real load and the action expected before equipment limits are reached. If air cooling remains for residual heat or mixed racks, include it in the failure sequence rather than treating liquid cooling as a complete replacement by default.

Sources: ASHRAE, National Institute of Standards and Technology

Canadian climates can support economizer hours, but the useful result depends on outdoor temperature, humidity, smoke, contaminants, equipment limits and the chosen heat-rejection system. Compare the actual weather file and control sequence for the site. Do not describe a cool province as free cooling without stating the equipment, hours, fan and pump energy, water use and backup mechanical cooling required.

Sources: Natural Resources Canada, ASHRAE

For water-using systems, identify source, quality, treatment, storage, discharge, seasonal limits, permits and interruption response. Separate water withdrawn from water consumed. Include chemical handling, drift, plume, freeze protection and maintenance. Compare water and energy together because a change that lowers one can raise the other. Keep public environmental claims within the measured boundary and reporting period.

Sources: Natural Resources Canada

Collect representative inlet temperature, humidity where required, supply and return conditions, differential pressure, flow, valve position, fan or pump speed, cooling power and IT power. Choose intervals that show control changes and short events. Set alarm priorities and delays so staff can distinguish a developing thermal problem from harmless noise. Keep sensor location, calibration and replacement records with the trend data.

Sources: Natural Resources Canada, ASHRAE

Review performance by load and season. A single annual number cannot show a stuck valve, simultaneous heating and cooling, excessive fan speed, failed containment or poor low-load control. Annotate major IT deployments, control changes, maintenance and outages. Natural Resources Canada recommends metering and performance tracking because improvement depends on seeing where energy is used and whether a change produced the expected result.

Sources: Natural Resources Canada

Build tests for loss of one cooling unit, pump, fan, control panel, sensor, power source and network path used by controls. Include utility transfer, generator operation, restart sequence, high outdoor conditions and the effect of a maintenance isolation. Measure the time and temperature response at representative racks. Set abort limits and restoration steps before the test begins.

Sources: National Institute of Standards and Technology, ASHRAE

After acceptance, keep operating ranges, alarm logic, setpoints, trend views, test records, maintenance tasks and escalation contacts under change control. Recheck the thermal model when rack density, server generation, liquid-cooling share or room layout changes. Efficiency work should protect equipment requirements and recovery behaviour. A lower energy number is not a success if it removes the margin needed for a defined fault or maintenance state.

Sources: Natural Resources Canada, National Institute of Standards and Technology