Topology

2N, N+1 and distributed redundancy — as statements you can test

Redundancy labels are shorthand for a promise about what the facility survives. The label is worth exactly as much as the drawing behind it, and the arithmetic that checks the label is not the arithmetic most people use.

The configurations, stated as what they survive

Two different promises, often confused

Redundancy labels answer one question; the design brief usually asks two.

• Concurrent maintainability — every capacity component and every element of the distribution path can be removed from service on a planned basis without touching the critical load. This is a statement about one component at a time, chosen by you, at a time of your choosing.
• Fault tolerance — the facility survives a single unplanned failure anywhere, including during maintenance, without touching the critical load. The event is not chosen by you and the timing is not either.

N+1 on its own buys the first and not the second: the spare unit covers the unit, but the bus they share is still one failure. This is the single most common gap between what a brief asks for and what a drawing delivers.

The arithmetic mistake in a distributed plant

Take four generator sets of 2,000 kW carrying 6,000 kW of load — a 4N/3 arrangement. The familiar check is pooled: total capacity 8,000 kW, lose one set, 6,000 kW remains, load is 6,000 kW, so it holds.

That check is only valid if every remaining set can actually reach every part of the load. In a distributed plant it usually cannot. The sets do not share one bus; each feeds a subset of the boards, and the dual-corded loads are split between them in a pattern the drawing fixes. When one set is lost, its share does not spread evenly across the other three — it lands on the specific sets that share its loads.

Both rows describe the same plant. The first is a sum; the second is the drawing. Only the second is a design check.

What to ask of any redundancy claim

1. Which components are counted as N, and at what load? A claim without a load is not a claim.
2. Is the spare on the same bus as the units it covers? If yes, the bus is the limit, not the spare.
3. Does the claim hold during maintenance, or only when everything else is healthy?
4. In a distributed plant: with each source lost in turn, what does every remaining source carry?
5. Does cooling carry the same label as power? A 2N electrical plant in front of an N chilled-water plant is an N facility.

Where cooling usually breaks the claim

Power is drawn, reviewed and argued about; cooling is often labelled by unit count alone. The questions that decide it are the same ones: can a chiller, a pump and a room unit each be taken out on a planned basis with the load carried? And if continuous cooling is claimed, are the chilled-water pumps on uninterruptible power — or is there a thermal store sized for the chiller restart? Room units on a UPS while the pumps are on the generator is not continuous cooling.