Data Centre Design
The Data Centre Design domain tests your ability to read a facility as an integrated physical system: site, building, power, cooling, cabling, fire and security. Exam questions typically ask which design choice supports a stated availability goal, or which single weak path defeats an otherwise redundant design.
Core design rules
- Site and building: evaluate flood risk, utility feeds, fibre routes, climate, physical security, expansion space, floor loading, ceiling height and loading-dock access. A technically strong MEP design can still fail on site hazards.
- Power chain: utility → MV/LV switchgear → transformer → UPS → PDU → rack PDU, with generators, ATS/STS and fuel systems. Learn redundancy topologies: N (no spare), N+1 (one spare component), 2N (two independent full paths) and 2(N+1). Concurrent maintainability means any path can be serviced without shutting IT load; fault tolerance means the load survives a single failure without downtime.
- Cooling: design around actual heat load, not room area. Use hot/cold aisle arrangement, containment, CRAC/CRAH units, chilled-water plant, heat rejection and economisers. Check the ASHRAE TC 9.9 recommended envelope and the cooling path’s redundancy separately from the power path.
- Cabling, fire and security: structured cabling with redundant pathways, separation from power, fibre and copper diversity; detection such as VESDA, clean-agent or water-mist suppression, compartmentation; layered access control and mantraps.
Worked example
A design has two utility feeds, 2N UPS and N+1 generators, but one central chiller plant with N chillers and no thermal storage. The client wants TIA-942 Rated-3. Verdict: the power path may be concurrently maintainable, but the cooling path is not. Losing a chiller, pump, valve or pipe section can force IT load reduction. Rated-3 requires every critical path, including heat rejection, to be concurrently maintainable at full load. The design is not Rated-3 unless the chiller plant and its distribution are made redundant. Reasoning: rate the weakest path, not the best component.
Traps that cost marks
- Treating N+1 as equivalent to 2N; N+1 has a single spare, while 2N has two complete independent paths.
- Assuming UPS redundancy compensates for a single generator or a single utility feed.
- Comparing nameplate capacity with usable capacity after derating, diversity and growth.
- Ignoring floor loading, commissioning and maintainability access during design review.
- Assuming a component rating automatically gives a facility rating.