N+1 and N+N get used interchangeably in sales conversations about power redundancy, and they shouldn’t be — they represent genuinely different failure tolerances, different costs, and different practical outcomes when something actually goes wrong at 3am. If you’re colocating anything business-critical, understanding the difference is worth ten minutes of your time.
Defining N first
“N” is simply the capacity actually needed to run the load — the number of UPS modules, generators, or cooling units required to support the facility’s current draw, with zero spare. An N-only system has no redundancy at all: if any single component fails, there’s an immediate capacity shortfall.
N+1: component redundancy
N+1 adds exactly one additional unit of capacity beyond what’s needed. If the facility needs four UPS modules to carry the full load, an N+1 configuration runs five — so any single module can fail (or be pulled for maintenance) without capacity dropping below what’s needed. This is the minimum sensible standard for any facility hosting production workloads, and it’s what underpins Tier II and Tier III classifications.
The limitation: N+1 protects against a single component failure, but the spare capacity is typically shared across a single distribution path. If that whole path — the switchgear, the busbar, the distribution board feeding it — has a fault, N+1 redundancy at the component level doesn’t help you, because the fault isn’t in a UPS module, it’s in the path itself.
N+N: full system redundancy
N+N (sometimes written 2N) doubles the entire system rather than adding one spare unit. If N is four UPS modules on one distribution path, N+N is a completely independent second set of four modules on a completely independent second distribution path — separate switchgear, separate busbars, ideally separate physical routing through the building. Either path alone can carry the full load. This is what makes true “A feed / B feed” dual-corded power possible at the rack: your server’s two power supplies each draw from an entirely separate, independently capable system.
N+N is what underpins genuine Tier III concurrent maintainability (see our guide to UK data centre tiers) and all of Tier IV fault tolerance. It costs more to build and operate than N+1 — you’re maintaining twice the infrastructure — which is reflected in what facilities running true N+N charge for power and space.
| N (no redundancy) | N+1 | N+N (2N) | |
|---|---|---|---|
| Protects against single component failure | No | Yes | Yes |
| Protects against distribution path failure | No | No | Yes |
| Concurrently maintainable | No | Partially | Yes |
| Relative capital/operating cost | Lowest | Moderate | Highest |
| Typical Uptime tier alignment | Tier I | Tier II | Tier III/IV |
Why this matters at the rack, not just the facility
Here’s the part that catches people out: a facility can have genuine N+N power infrastructure right up to the cabinet, and your deployment can still have a single point of failure — because your server only has one power supply, or because both of its cords are plugged into the same PDU strip instead of one into each A and B feed. N+N protects you only if your own hardware is actually wired to use both independent paths. Dual power supply units (dual-corded servers) with each cord terminated on a separate feed is the piece that has to happen at your end.
What to actually ask for
- Confirm whether power redundancy is N+1 or genuine N+N at the facility level, not just at the UPS module level.
- Ask whether your specific rack gets two physically independent feeds (A and B) or one feed with a redundant module upstream.
- If your servers are dual-corded, confirm during setup that each cord is terminated on a separate PDU tied to a separate feed — this is easy to get wrong during a rushed install.
- Ask how generator and UPS failover is tested, and how often — a redundant design that’s never load-tested is a theoretical one.
Our colocation racks run on genuine N+N power distribution with independently routed A/B feeds to every cabinet, so dual-corded deployments get real path-level redundancy rather than component-level redundancy alone. If you want to check your current or planned deployment is actually wired correctly, our team can talk you through it.
