Data Centre Operations and Maintenance
Operations and Maintenance asks a different question from design: not what the facility could deliver, but what it actually delivers on a Tuesday afternoon when a UPS module fails. The exam tests procedural control, documentation discipline and maintenance planning.
The three procedure types
- SOP - Standard Operating Procedure: a routine, repeatable task performed in a stable environment, such as a generator load test.
- MOP - Method of Procedure: a specific, non-routine change with a step sequence, dependencies, hold points, verification and rollback, such as replacing a UPS module.
- EOP - Emergency Operating Procedure: the scripted response to an unplanned event, such as total loss of chilled water.
Change and work control
Change request → risk assessment → change advisory board approval → scheduling in a maintenance window → permit to work → lockout/tagout with verification of zero energy → execution against the MOP → functional verification → drawing and CMMS update → close-out and post-implementation review.
Maintenance strategy
Reactive (run-to-failure, acceptable only for non-critical items), preventive (time or usage based) and predictive or condition-based (thermography, vibration analysis, oil analysis, battery impedance testing). The CMMS holds the asset register, work orders and maintenance history; spares holdings and supplier lead times drive MTTR far more than technician skill alone.
Monitoring and capacity
BMS covers mechanical and electrical plant, EPMS covers power metering and power quality, DCIM aggregates assets, capacity and connectivity. Alarms need thresholds, escalation paths and owners, otherwise they become noise. Capacity management tracks rack power, cooling, floor weight, ports and space.
Worked example
Replace one module in an N+1 UPS serving a Tier III hall. First establish that the hall load is below the capacity of the remaining N modules - say the load is 60% of N capacity, so no risk. Then confirm the MOP is approved, isolate the module with lockout/tagout, prove dead with a tester, execute the swap at the defined hold point, restore, verify output and alarms, and record the work order. Rollback is stated in advance: re-energise the original module and restore the original configuration.
Traps that cost marks
- Swapping SOP and MOP labels.
- Submitting a change with no rollback step or verification step.
- Lockout/tagout without proving the circuit dead.
- Treating DCIM as a substitute for BMS/EPMS control.
- Scheduling preventive work during a change freeze or peak period.
- Recording maintenance in email instead of the CMMS.