UPS redundancy: N, N+1 and 2N by three-phase model

The five designs, from N to 2N

Schneider Electric's White Paper 75 compares the five UPS designs most often built, and ranks them on a scale of availability from 1 (lowest) to 5 (highest). Its availability figures come from a simplified model that, among other things, leaves out human error, so they read higher than a real installation; they are for comparing designs, not for promising uptime.

How N+1 works in practice

In a parallel redundant system the modules must be identical in capacity and model. Their outputs are synchronized, they share the load evenly, and when one leaves the bus for service or after an internal fault, the others take its load at once, without sending the critical load to utility power. The system is N+1 when the spare power equals at least one module, and N+2 when it equals two.

WP 75 works an example: a room whose design load is 400 kW is an N system with one 400 kW UPS or two 200 kW UPS in parallel. To become redundant it needs two 400 kW modules or three 200 kW modules on a common bus, and that bus is sized for the non-redundant capacity: 400 kW. N+1 also leaves room to grow, if a new module is ordered when the load reaches a set share of the installed capacity.

Redundancy does not remove every weak point. The paralleling bus remains a single point of failure, upstream and downstream equipment can still fail, and service on the parallel board or controls exposes the load. That is why N+1 systems keep a static bypass and, in most cases, a maintenance bypass; see maintenance without shutdown, N+1 vs 2N.

Module size and efficiency: a 240 kW example

Lightly loaded UPS modules are usually less efficient than modules loaded closer to capacity, and how much varies by manufacturer. WP 75 shows how the module size chosen for N+1 sets the normal load on each module, for the same 240 kW load:

Smaller modules keep each one more heavily loaded and the spare smaller, at the cost of more modules to maintain.

Redundancy and parallel limits by model

Sources, checked October 3, 2026: values as each manufacturer publishes them for the model shown; “Not stated” means the page we read does not give the value.

N+1 by model

The table above lists each family's published redundancy and parallel limit; the modular vs monolithic guide explains the frames. Three model points: whole units, modular or monolithic, also reach N+1 by paralleling, up to the maker's limit. Eaton lists the 9395XC at 2 units maximum (3 MW), so a 9395XC system built as N+1 carries the load of one unit. The 9395XR is built from 125 kW power modules, 12 per unit, and parallels up to 4 units (6,000 kW).

Isolated redundant is not the same N+1

An isolated redundant system is sometimes called N+1, but it works differently: a secondary UPS sits unloaded and feeds the static bypass of the primary. It needs no paralleling bus and the two units need not match, which is why it is used to add redundancy to an existing UPS without replacing it. The secondary must take the full load in one step when the primary transfers to bypass, after running at 0% for a long time, and not every UPS can.

WP 75 cites a probabilistic risk assessment by MTechnology: both an isolated redundant and a non-redundant system showed an unreliability of 1.8% per year, while the isolated design had 30 failure modes against 7, and more room for human error.

2N and distributed redundant

In 2N, each path carries the full load on its own, so the parallel limits above apply to each side separately. In the 2(N+1) example in WP 75, a 300 kW load is served by four 300 kW modules, two on each of two parallel buses. Downstream, dual-corded equipment takes both paths directly; single-corded loads need a transfer switch, and WP 75 notes that a static transfer switch moves the load to its other source in about 4 to 8 milliseconds. See the bypass and STS guide and the data-centre redundancy paper.

Distributed redundant designs give concurrent maintenance, when all loads are dual-corded, with fewer modules than 2(N+1). For a 3.2 MW data centre, WP 75 counts ten 800 kW modules for 2(N+1), five per bus, against six 800 kW modules for distributed redundant. The trade: designs with STS depend on them, and WP 75 lists complex failure modes and many operating modes that are hard to test.

Redundant designs are maintained under data-centre UPS service, and modular frames that support N+1 are listed among the three-phase UPS families online.

Why the UPS layer matters

Uptime Institute's Annual Outage Analysis 2026 finds power still the leading cause of impactful outages, with failures of UPS systems, transfer switches and generators dominant. Its 2026 resiliency survey puts UPS failures first among the causes of major power-related outages over the past three years, followed by transfer switch and generator issues, and reports that power issues caused more than one in five IT service outages in that period.

Choosing a design

• Cost of downtime: what an hour of outage costs, and how long recovery takes, drives the budget discussion.
• Maintenance windows: if the business can shut down every night and most weekends, WP 75 says the design need not go far beyond parallel redundant; the less time a year allows for maintenance, the more redundancy the system needs.
• Loads: dual-corded loads let a design use two paths fully; single-corded loads need transfer switches close to them.
• IT architecture: with virtualization and fast networks, WP 75 expects two remote data centres with 1N redundancy to be more available than one highly redundant data centre.
• Budget: 2(N+1) costs significantly more than capacity, parallel redundant or distributed redundant designs.

Frequently asked questions

What is N+1 UPS redundancy?

N+1 means one more UPS unit or power module than the load needs, so any single unit or module can fail or be serviced while the rest carry the load.

When do I need N+1?

When a single UPS module failure, or service on one module, must not put the load on unprotected utility power. Work on the paralleling bus or controls still exposes it. If the site can run on bypass during service, N may be enough.

How many Eaton 9355 units can run in parallel?

Eaton lists paralleling of up to four 9355 UPSs, for capacity or redundancy, with its HotSync technology.

Can one APC Galaxy VL frame be N+1?

Schneider lists the GVL200K500DS, 200 kW scalable to 500 kW, as redundant, with 4 power modules installed and 6 free slots.

How many Liebert EXL S1 units can be paralleled?

Vertiv lists up to 8 units, in distributed or centralized parallel configurations.

What is 2(N+1) redundancy?

Two complete parallel redundant (N+1) UPS systems feeding two paths. A single module failure is absorbed inside its own system, so UPS redundancy remains even during concurrent maintenance.

Is isolated redundant the same as N+1?

No. Isolated redundant uses an unloaded secondary UPS that feeds the primary's static bypass; parallel redundant N+1 shares the load across identical modules on one bus.

Related guides and systems

Request a system quote

Tell us the application and we reply within 4 business hours with the sizing, the right system, the installation and a written quote. Three-phase installs usually need a licensed electrician, so let us know if you have one.