Data Centre Operations and Maintenance
The Data Centre Operations and Maintenance domain tests day-2 control: how work is planned, executed, monitored and recorded so that availability and safety are protected. Expect questions on procedures, maintenance, change control and performance measurement.
Operational control
- Documented procedures: MOPs for methodical work, SOPs for routine operation, EOPs for emergencies. A change should have a risk assessment, approval, implementation window, rollback plan and post-change verification.
- Safe work control: permit-to-work, lockout/tagout, access control and escort rules, especially for UPS, switchgear and generator work.
- Preventive maintenance: UPS battery inspection and capacity testing, generator load tests and fuel testing, ATS transfer tests, chiller and pump maintenance, infrared thermography, breaker and torque checks.
- Monitoring: BMS/EMS and DCIM for power, cooling, humidity, leak detection, alarms and capacity. Alarm thresholds and escalation paths must be defined.
- Capacity management: track power, cooling, space, network ports and weight; use actual versus design capacity and forecast growth.
- Incident and problem management: severity classification, escalation, communication, root-cause analysis and corrective actions. Track SLAs and KPIs such as availability, PUE, MTTR and MTBF.
Worked example
A UPS battery string fails a capacity test and delivers only 80% of its design runtime. The operator wants to defer replacement because IT load is currently 40% of design. Reasoning: the battery is degraded; the relevant question is whether remaining autonomy at actual load still covers generator startup and safe shutdown under the approved risk. The operator should record the finding, assess risk, apply temporary controls such as load restriction or a spare string, schedule replacement, and update the change and maintenance records. Deferral is a risk decision, not a paperwork decision.
Traps that cost marks
- Testing a generator only for startup with no load; this does not prove it can carry IT load.
- Treating PUE as an efficiency percentage rather than a ratio of total facility power to IT power.
- Skipping change records for “minor” work that still affects critical paths.
- Confusing preventive maintenance with predictive condition monitoring.
- Measuring availability without agreeing the measurement boundary and downtime definition.