Redundancy topologies [/rɪˈdəndənsi ˈtopologies/] n - Redundancy is the tongue in which we say how many spare components a power system holds, and how they are arrayed. It is the single weightiest design decision in The Engines of the Hall, for it settles what may fail (or be taken offline for maintenance) without letting fall the load.

The Vocabulary of It

• N [/ˈn/] n - capacity exactly sufficient to the load, and no more. One failure lets something drop.
• N+1 [/ˈn1/] n - one spare beyond what the load requires. Any single component (a The Standby Engine, a Uninterruptible power supply module) may fail or be serviced whilst the rest bear the load.
• 2N [/ˈ2n/] n - two whole and independent systems, each sufficient to 100% of the load. An entire path (utility feed, UPS, Power distribution unit, Busway distribution) may be lost and the site stands. It is this that feeds the A/B cords of a Rack power distribution unit.
• 2N+1 [/ˈ2n1/] n - two full systems, and a spare in each besides. Belt and braces.

Maintainable Against Tolerant

• Concurrently maintainable [/ˈkɒnkʌrrɛntli ˈmæɪntæɪnæblɛ/] n - any component may be drawn out for planned maintenance without downtime. Oft achievable with N+1.
• Fault tolerant [/ˈfæʌlt ˈtɒlɛrænt/] n - any component may fail unlooked-for without downtime. This generally asks 2N.

These two properties are the very thing the Uptime Institute tier classification grades. Their attainment hangs upon truly independent paths; which is why Static transfer switch and Automatic transfer switch devices appear wherever two such paths must join upon one load.

Part of The Engines of the Hall.