Data centre redundancy: N, N+1 and 2N

N is the amount of equipment needed to carry the load. N+1 adds one spare unit so a single failure does not drop the load; 2N builds two complete, independent systems so a whole path can fail or be shut down for maintenance. The right level depends on what an outage costs you, whether your equipment is dual-corded, and how you plan to maintain the room.

What N means

N is the capacity the load needs, counted in units of equipment. If the IT load and its supporting loads need 300 kW of UPS capacity and the UPS is built from 100 kW modules, N is three modules. The same counting applies to generators, cooling units, pumps and anything else the room depends on.

A system built to N has no spare. It works as long as every unit works. When one fails or needs servicing, the remaining units are overloaded, and the UPS either transfers to bypass (raw utility power, with no battery behind it) or the load drops. N is acceptable for rooms where a planned shutdown for maintenance is tolerable and an occasional unplanned outage is survivable.

N+1, 2N and the levels between

Each step up the ladder adds equipment, and with it cost, space and losses. What it buys is the ability to lose something without losing the load. The diagram compares the three common levels using UPS modules as the example.

N+1

N+1 is the most common level for UPS modules, generators and cooling units. With four 100 kW modules carrying a 300 kW load, any one can fail and the other three still carry it. Each module runs at about 75 percent of its rating in normal operation. For larger systems, N+2 adds a second spare so that one unit can be out for maintenance while another fails.

The limit of N+1 is that redundancy usually stops at the modules. They all feed one output switchboard, one set of distribution and one cord to each server. A fault in that shared equipment, or a mistake while working on it, can still drop the load.

2N

2N builds two complete systems, usually called A and B, each sized for the full load. Dual-corded servers take one cord from each. In normal operation each side carries roughly half the load, which means each side must be kept below half of its capacity so it can pick up everything if the other side is lost. That is the cost of 2N: twice the equipment, each piece lightly loaded.

2N+1 adds a spare module within each side, so one side can be out for maintenance while the other still has a spare.

Distributed redundancy

Some larger facilities use designs that sit between N+1 and 2N, such as three UPS systems sharing the load of two, with each server's two cords connected to different pairs. These designs save equipment compared with 2N but need careful load planning: a rack added to the wrong pair of systems can quietly remove the redundancy everyone thinks is there.

Where single points of failure hide

Counting modules is the easy part. The question that matters is whether there is any single piece of equipment, cable or breaker whose failure drops the load. In many N+1 systems, the answer is yes, and it is downstream of the redundant part.

Other common single points include a single utility service, a single generator feeding both sides, a single fuel system, one cooling plant with one set of pumps, a shared control system that can command both sides at once, and the rack itself if equipment is single-corded. A thorough redundancy review traces each of these, not just the UPS.

Concurrent maintainability and fault tolerance

Two ideas sit behind the tier classifications used in the industry. A concurrently maintainable system lets any component be shut down for planned maintenance without dropping the load. A fault-tolerant system survives any single unplanned failure, including one that happens while something else is out for maintenance, without dropping the load.

The Uptime Institute's Tier I to IV classification and the rated levels in ANSI/TIA-942 both build on these ideas. Roughly, the lower levels describe basic capacity and redundant components, the third level requires concurrent maintainability, and the top level requires fault tolerance. They are useful shared vocabulary, but the details and any certification process belong to those organisations, so check the current documents before committing to a target in a contract.

Single-corded equipment

2N only protects equipment that can use both feeds. Older servers, network appliances and some storage have one power cord. A rack-mounted automatic transfer switch, or a static transfer switch upstream, connects that equipment to both sides and switches to the surviving side if one fails.

Transfer switches are themselves a single point of failure for the equipment behind them, so they are a stopgap rather than a substitute for dual-corded equipment. Keep an inventory of single-corded devices and plan to replace them over time.

Choosing a level

1. Put a cost on downtime. Lost revenue, lost data, contract penalties and reputational harm per hour of outage. This is the number every other decision is weighed against.
2. List what the room depends on. Utility, generator, fuel, UPS, distribution, cooling, controls and network. Redundancy in power without redundancy in cooling buys little, because the room overheats instead.
3. Decide how maintenance will happen. If the business cannot accept an annual shutdown, the design must be concurrently maintainable.
4. Check the IT equipment. Count dual-corded and single-corded devices. 2N without dual cords is mostly wasted.
5. Write down the loading rules. In 2N, each side stays below half its capacity. In N+1, the total stays below the capacity of N units. These limits must be monitored, because growth erodes redundancy quietly.

Generators follow the same logic. A single generator is a single point of failure; two or more paralleled generators with N+1 capacity are common for larger sites. Our guide to generator paralleling explains how they share load, and the guide to UPS bypass and maintenance covers how a UPS is serviced without dropping load. For the full sequence of equipment, see the data centre power chain.

How ehZee can help

We review existing rooms for single points of failure and loading against redundant limits, and design new or expanded power systems at the redundancy level the business chooses. Drawings for permit are sealed by a licensed professional engineer where required, and installation is done by licensed electrical contractors under permit. See our data centre electrical design service.

Common questions