Data centre design
Domain 1 - Data centre design. In this domain the exam tests whether you can convert a business requirement (availability target, IT load, growth, budget, location) into a coherent facility concept - power topology, cooling architecture, space and cabling, fire suppression and physical security - and then defend the trade-offs numerically.
Start from the availability target, not from equipment
- N, N+1, 2N and 2(N+1) describe redundancy of the whole path. A 2N UPS feeding a single PDU, or 2N power with one cooling loop, is not 2N.
- Concurrently maintainable (Tier III / rated-3): every component, including the cooling path, can be maintained on a planned basis without shutting down IT load.
- Fault tolerant (Tier IV / rated-4): the facility survives a single unplanned fault and remains concurrently maintainable.
- Redundancy must match across power, cooling and network paths; the weakest link defines the real class.
Size the electrical chain, then the thermal load
- UPS capacity is specified in kVA at a stated power factor. Diversity factors and growth headroom are design decisions that must be written down, not assumed.
- PUE drives electrical demand (facility kW = IT kW × PUE) but not the heat to be removed - the room cooling load is essentially the IT load.
- Support the topology with containment, hot/cold aisle discipline and an ASHRAE TC 9.9 envelope, or the redundancy stays theoretical.
Worked example
Brief: 400 kW IT, single site, Tier III, design PUE 1.4.
- Facility electrical demand = 400 × 1.4 = 560 kW.
- UPS: 400 kW at 0.95 PF = 421 kVA → N+1 = 2 × 500 kVA (not 2 × 450; 500 kVA leaves growth margin).
- Cooling: roughly 400 kW of heat plus fan and coil gains → about 420–440 kW installed capacity, N+1 CRAHs with isolation valves.
- Tier III requires the cooling path to be maintainable without load shutdown: dual-corded IT, A/B power, no single PDU feeding both feeds.
Traps that cost marks
- Equating Tier III with a fixed uptime percentage; the tier defines topology, not an SLA figure.
- Forgetting power factor and UPS derating, or sizing on nameplate rather than measured load.
- Designing 2N power but a single chilled-water loop or single chiller.
- Presenting design PUE as though it were measured performance.
- Ignoring airflow management (bypass, recirculation), which wastes the redundancy you paid for.