ENGINEERING NOTE / REDUNDANCY

Count redundancy by what it survives.

N+1 and 2N describe how much equipment is installed. The useful question is which events the arrangement can absorb, including the ones that arrive together.

Low Kah Yi · 4 October 2026 · Independent synthetic example

N is the load, not the equipment.#

Redundancy notation starts from the load. N is the number of units needed to carry it. If a load needs three cooling units, N+1 installs four, so any one can be unavailable. 2N installs two complete groups of three, each able to carry the whole load on its own. 2(N+1) gives each of those groups its own spare.

Those labels count equipment. They do not say what happens when something goes wrong, and that is the part a design review needs. A clearer way to compare arrangements is to list the events that matter and ask, for each one, whether the load stays fully served.

Read the table by its columns.#

Every arrangement except N survives a single unit fault, and the same is true of a single unit in maintenance. The difference appears when events combine. With N+1, maintenance uses the spare, so a fault during the work interrupts the load. With 2N, the second group absorbs both. That is the practical meaning of being able to maintain a system while it keeps running: planned work and an unplanned fault can overlap.

The last column is the one most often missed. With 2N, if maintenance takes a unit out of one group and the other group then loses its supply path, two units remain for a load that needs three. Giving each group its own spare, 2(N+1), is what closes that gap. Each additional column survived is a deliberate choice with a cost.

Count the paths, not just the units.#

The table assumes each group has its own distribution path. The third row shows what happens when it does not: six units on one shared path survive every unit-level event and still stop together when that path fails. The equipment list says 2N; the arrangement behaves like a single system for that dependency.

Shared paths are easy to overlook because they are rarely equipment you count: a common loop, a single switchboard, one control network or one monitoring system. Dependencies are part of the design follows one of them through a synthetic example.

Try it: in the Resilience Studio, compare the shared and independent loop arrangements with one cooling unit in maintenance, then add a second fault.

N moves with the load.#

Redundancy is defined against a load, so it can be lost without any equipment changing. If the load grows until four units are needed, the N+1 arrangement above becomes N, and its spare is already in use. Margin reviews belong with capacity planning for that reason. Capacity margin determines what happens after restoration shows the same effect in the thermal response.

Choose the events, then the arrangement.#

The right arrangement depends on which combinations of events the service has to ride through, how often planned work happens and how long a repair takes. Writing those events down first turns a debate about equipment counts into a comparison of what each option survives, and what it costs.

This note describes general engineering practice. The counting model treats units as interchangeable and ignores partial capacity, switching time and common-cause failures. It does not establish a redundancy classification, availability figure or design for any facility. Read its source on the model code page.