Data Centre Operations and Maintenance
This domain tests how design availability is preserved in service: procedures, change control, maintenance, monitoring, capacity management, and incident response. Questions often present an operational task and ask whether it needs a routine procedure, a planned change, or an emergency response.
Operational framework
- SLAs define service commitments; OLAs and vendor contracts support them. Availability depends on MTBF and MTTR, so fast, correct recovery matters.
- Documentation set: as-built drawings, rack elevations, cable schedules, SOPs, MOPs, and EOPs.
- SOP = routine, repeatable task. MOP = controlled method for a non-routine or risk-bearing change. EOP = response to an unplanned event or loss of redundancy.
- Change management: request → risk assessment → approval → scheduled window → implementation → verification → documentation update, with a rollback plan.
- Maintenance includes preventive, predictive, and condition-based work: battery testing, generator load tests, filter changes, torque checks, and infrared scans.
- Monitoring uses BMS/DCIM alarms, thresholds, escalation, and trend analysis. Capacity planning covers power, cooling, space, and network.
- Safety controls include lockout/tagout, PPE, arc-flash boundaries, and permit systems.
Worked example - replacing one UPS module in an N+1 system. This is not a routine SOP because it changes the electrical configuration and temporarily removes redundancy. It needs a change request, risk assessment, and MOP. The MOP should confirm the remaining module can carry the load, verify generator/ATS readiness, notify stakeholders, isolate the module under LOTO, replace it, test transfer and load sharing, then update drawings and records. If load is dropped or redundancy is lost unexpectedly, the EOP governs. Reasoning: the task is planned, but the risk and configuration change make it a controlled change, not a checklist SOP.
Traps that cost marks
- Using an EOP as a substitute for a planned MOP, or an SOP for a configuration change.
- No rollback plan or no post-change verification.
- Deferring preventive maintenance because no alarm has occurred.
- Capacity planning based only on current load, ignoring growth and redundancy loss.
- Treating an alarm as a resolved incident without root-cause follow-up.