UPS Redundancy: N+1, 2N and Why It Matters

This guide explains the redundancy configurations with Schneider Electric's published definitions, shows what N+1 costs in capacity, and works through a real rack using APC units sold here, with prices, plugs and the failure each design does and does not survive.

Written by the UPSPLUSBATTERY technical team. Sources checked 25 September 2026.

Scope and terms

This article covers UPS redundancy for server rooms, network closets and small data rooms, from a single rack to a modular three-phase frame. Generator, utility and cooling redundancy are out of scope.

• N: the capacity the critical load needs. A 240 kW load needs N = 240 kW of UPS.
• N+1 (parallel redundant): enough modules to carry the load, plus one, all feeding a common output bus.
• 2N (system plus system): two complete UPS systems, each sized for the whole load, feeding separate distribution paths. It is a facility design, not just two boxes.
• Dual-UPS (A/B) at the rack: the small-site version: two UPS, each able to carry the rack, one feeding each cord of dual-corded equipment.
• Dual-corded: equipment with two power supplies and two cords. Check that each supply can carry the device alone (often listed as 1+1).
• VA and W: UPS carry two ratings. Size on watts (W), the real power your equipment draws; the VA figure is always equal or higher.

The five configurations and what each protects against

Schneider Electric's White Paper 75 ranks the five common UPS designs on a scale of availability from 1 to 5 and maps them to data-centre tiers. Table 1 summarizes it. The cost column is the paper's whole-facility build-out cost per rack (generators, switchgear, cooling, fire suppression, space and the UPS), in US dollars at the time of Revision 3. It is not a UPS price and not current pricing.

For a whole facility, then, system plus system cost the paper about 2.3 to 2.7 times a capacity-only design (36,000 / 13,500 = 2.67 at the low end, 42,000 / 18,000 = 2.33 at the high end). Two cautions. The tiers belong to the Uptime Institute, which defines them by redundant capacity components and distribution paths across the whole facility, not by the UPS alone. And White Paper 75 warns that an isolated redundant system is sometimes called N+1 but is considerably different from a parallel redundant one, so ask which one a quote means. Schneider's current edition of the paper is Version 4; the figures above are from Revision 3.

What N+1 costs in capacity

In N+1 every module carries part of the load, so none of them runs near its rating. White Paper 75 shows the effect for a 240 kW load (Table 2): more, smaller modules leave less capacity idle, but add more parts that can fail, and the paper notes that lightly loaded modules are typically less efficient.

The same paper points out what N+1 does not fix: even with a spare module, the paralleling bus that joins the modules remains a single point of failure. Figure 1 compares it with the dual-path approach.

Redundancy at rack level: two UPS and a transfer switch

Most small sites do not need a parallel UPS system. If the servers are dual-corded, feeding cord A from one UPS and cord B from another means either UPS can fail or be switched off for service while the other carries the rack. The two paths are from the UPS inputs to the equipment, as long as the UPS sit on different circuits; a transfer switch is the one place where they join.

The weak point is equipment with one cord. Schneider Electric's White Paper 62 notes that single-corded equipment still makes up about 10 to 20% of IT equipment in mission-critical facilities, and, based on its selection criteria, names a rack-mount automatic transfer switch (ATS) as the optimal solution below 10 kVA. The timing works: the same paper puts electromechanical transfer at 8 to 16 milliseconds, and cites the ITIC curve, under which IT equipment is designed to keep operating for 20 milliseconds at zero volts. APC lists a typical transfer time of 10 ms for the APC AP4450A rack ATS (120 V, 15 A, 10 NEMA 5-15R outlets), so size what you plug into it within 15 A. For 230 V racks, the AP4421A (230 V, 10 A, C14 inlet, 12 C13 outlets) and AP4423A (230 V, 16 A, C20 inlet, 8 C13 and 1 C19 outlets) fill the same role. White Paper 62 also warns that a transfer switch has its own failure modes, including its output breakers, so it protects the single-corded device from a UPS failure but not from its own.

Worked example: a 1,200 W rack

Take a rack drawing 1,200 W, measured on the UPS display or a metered PDU, all dual-corded except a firewall and a switch. Two ways to protect it with APC Smart-UPS On-Line rack units sold here:

Option B adds a second unit and, for single-corded devices, a rack transfer switch; check current prices on the product pages. What that buys is that the dual-corded servers stay on conditioned UPS power through a UPS failure or a UPS service. Four conditions make it real: put the two UPS on different circuits, ideally from different panels; check that each server's supplies can each carry the server alone, because some run active and standby rather than sharing; size for the failover state, not the normal one (89% leaves little room to grow, and in field practice designers often keep each unit nearer 80% in that state); and check runtime for one unit carrying everything with the UPS runtime calculator, adding external battery packs if a few minutes is not enough. Ratings, plugs and runtimes come from the product listings for the SRT1500RMXLA and SRT3000RMXLA; both are online double-conversion units, and if you are still choosing the topology, see line-interactive vs online UPS. The full range is in the APC Smart-UPS SRT collection. Give each UPS a network card and shutdown software, so servers still shut down cleanly if both paths are lost.

Modular N+1 inside one frame

Above a few kilowatts, modular UPS put N+1 inside a single frame: power modules share the load, and one more module than needed is installed. APC's Symmetra PX 40 kW three-phase modular UPS works this way. On the older single-phase Symmetra LX, a failed power module can be replaced while the others carry the load; UPSPLUSBATTERY stocks a refurbished SYPM4KP 4 kVA (3.2 kW) power module, which is not new and not supplied by Schneider Electric; follow Schneider's module replacement procedure for your frame. Table 2 applies to sizing: with four modules installed for a load equal to three modules' capacity, each runs at 75% of its rating. Keep White Paper 75's caution in mind: the frame, backplane, bypass and output are shared, so modular N+1 protects against a module failure, not a frame failure.

Maintenance is where redundancy pays

White Paper 75 notes that in a capacity-only (N) system the load is exposed to unprotected power during system maintenance, usually at least once a year for 2 to 4 hours. With two UPS at the rack, one unit can be taken out for replacement, electronics service or a firmware update while the other carries the rack. With a single UPS, that work runs on an external maintenance bypass, if one is fitted, with the load on raw utility power for the duration; our service partner explains what that means in its guide to UPS bypass mode and when it is critical. Plan battery replacement per unit: each SRT1500 takes its own pack, such as this compatible replacement battery pack for the SRT1500XLA (not an APC-made part; confirm the fit for the RM model). For multi-unit sites, our service partner' UPS preventive maintenance service schedules the work one path at a time.

Decision checklist

1. Measure the load in watts, from the UPS display or a metered PDU, not the nameplate total.
2. Count single-corded devices. If there are any, plan a rack ATS for them and accept it as their single point of failure.
3. Choose the failure you must survive: a module (N+1), a UPS (dual UPS at the rack), or a whole path including panels (2N facility design).
4. Size each unit to carry the whole load alone, and check runtime in that state.
5. Wire it: cord A of every dual-corded device to UPS A, cord B to UPS B, ATS input A to UPS A and input B to UPS B, each UPS on its own circuit. Label both ends.
6. Test it: switch one UPS off in a maintenance window and confirm nothing drops, then repeat with the other.

Sources

1. Schneider Electric, White Paper 75, Comparing UPS System Design Configurations, K. McCarthy and V. Avelar (current edition V4; figures quoted from Rev 3). Table 1, module loading table, efficiency note, paralleling bus, isolated redundant vs N+1, maintenance exposure of N systems.
2. Schneider Electric, White Paper 62, Powering Single-Corded Equipment in a Dual Path Environment, Rev 1, V. Avelar. Share of single-corded equipment, rack ATS recommendation below 10 kVA, 8 to 16 ms transfer, ITIC 20 ms, transfer switch failure modes.
3. Uptime Institute, Tier Classification System. Tier definitions.
4. Schneider Electric, AP4450A rack automatic transfer switch, product page. 10 ms typical transfer time.
5. Schneider Electric, Smart-UPS SRT3000RMXLA, product page. 3 kVA / 2.7 kW double conversion.
6. UPSPLUSBATTERY product listings for the SRT1500RMXLA, SRT3000RMXLA, AP4450A, AP4421A, AP4423A, Symmetra PX 40 kW and SYPM4KP (ratings, plugs, runtimes and prices as listed on 25 September 2026).
7. ThreePhaseUPS, Data centre UPS redundancy: N+1, 2N and distributed. Further reading for three-phase and data-centre scale designs.

APC, Smart-UPS and Symmetra are trademarks of Schneider Electric. UPSPLUSBATTERY is a reseller, is not affiliated with Schneider Electric, and Schneider Electric has not reviewed or endorsed this article. Compatible and refurbished parts are not manufacturer parts.