If a critical room has a single precision air conditioner, every failure and every maintenance of that unit is a moment of risk. Redundant operation is the name of the setup built to remove this risk. The most common method is N+1: one spare is added to as many units as needed.
However, there is a situation we often see in the field: the room has two air conditioners, but they cannot meet the load unless both run together. That setup is not redundant, it simply has two units. Below we explain step by step how to set up N+1 correctly.
What is N+1 redundancy?
N+1 redundancy means keeping one spare unit in addition to the N units that can carry the room's heat load on their own. N is the number of units needed to meet the load. For example, if two units meet the load, an N+1 setup has three units in the room, and when any one of them goes out of service, the remaining two are sufficient.
In more critical facilities, a 2N setup is also used: an exact copy of the whole group of units that carries the load is installed a second time. 2N provides higher security but the investment and space requirement also increase noticeably. For most corporate server rooms and mid-sized data centers, N+1 is a balanced solution.
The purpose of redundancy is not only protection against failure. In precision air conditioners, tasks such as filter replacement, condenser cleaning and humidifier maintenance require the unit to be stopped briefly. In a room with no spare unit, these maintenance tasks are either postponed or done at midnight, taking a risk. With an N+1 setup, maintenance can be done during the day, without interrupting the room's cooling at all, and without hurry.
How is capacity calculated?
The first step is to determine the room's total heat load: IT load, UPS and distribution losses, lighting, heat from the building envelope and a growth allowance. Then this load is divided by the sensible capacity of the selected unit under real operating conditions.
01Use the capacity under real conditions
In precision air conditioner catalogs, capacity is given for a specific return air temperature and outdoor condition. In climates with hot summers and cold winters such as Ankara, the unit's capacity on the hottest day should be used as the basis. Sensible capacity should be looked at rather than the total capacity in the catalog.
02Divide the load among the units
For example, if the total sensible load is 50 kW and the selected unit gives 28 kW sensible capacity at the design condition, two units are needed to meet the load (N=2). For N+1, a third unit is added. In normal operation, the three units run together at part load; when one stops, the other two go to full load.
Do the redundancy calculation not with today's load but with the load expected a few years from now. As cabinets are added, an N+1 setup can quietly slip to N+0, that is, to a state without redundancy. The load should be rechecked every year.
How should the units be run together?
In a redundant system, the units should not run independently of each other; they should be connected with a shared control network. Most precision air conditioners have a network feature that lets the units in the same room communicate with each other. This connection performs three basic functions.
- Rotation: The standby unit is changed at set intervals, so all units reach similar operating hours and the standby unit does not sit idle for years and turn out faulty when needed.
- Automatic switchover: When a running unit raises an alarm, the standby unit starts automatically.
- Shared control: The units work to the same temperature and humidity values; conflicts such as one humidifying while another dehumidifies are prevented.
In some facilities, running all units together at part load is preferred. With variable-speed fans and inverter compressor units, this method is generally more efficient, because fans use noticeably less energy at low speed. Which method to choose is determined by the unit type and the room's airflow.
Is redundancy limited to the air conditioner alone?
No. Installing three air conditioners and powering them all from the same fuse ties the redundancy to a single point. In a true N+1 installation, the points where a single failure could stop all cooling must be reviewed one by one.
- Electrical supply: Units should be fed from separate lines, preferably from different panels; restart of the air conditioners should be tested during generator changeover.
- Control network: A break in the communication cable should not stop all the units; the units should be able to keep running independently when communication is lost.
- Outdoor units: If condensers are placed crowded at the same spot, they draw in each other's hot air and all go into a high-pressure alarm together in summer.
- Chilled water supply: For CRAH type units, redundancy should also be considered on the chiller, pump and valve side.
How is a redundant system tested?
Is your server room really redundant?
We present our N+1 recommendation in writing, based on load measurement and single point of failure analysis.
The only way to know a redundant system works is to test it under controlled conditions. On a planned maintenance day, one of the units is taken out of service and it is monitored whether the remaining units keep the cabinet inlet temperatures within the recommended range.
During the test, how long the standby unit takes to start, whether the alarm reaches the notification system and how much the temperature rises should be recorded. This test should be done at least once a year, preferably before the summer months. For temperature targets, see our article on server room temperature.
Because these tests require work on the electrical panel and the control board, they should be performed by a qualified technician in coordination with the facility's IT team.
What to record in the test
- The unit taken out of service and the time of the test
- The standby unit's start-up time
- The highest temperature reached at the hottest cabinet inlet
- The time the alarm took to reach the notification system and the people it reached
- Current, pressure and fan values of the remaining units
When these records are compared with next year's test, an increase in the room's load or a loss of capacity in the units is noticed early. In this way, redundancy is tracked not on paper but as a measured fact.
How can redundancy be achieved in your existing room?
In an existing room, the first step is to measure the load and the real capacity of the units. Sometimes adding a new unit is enough; at other times, by adjusting the airflow, one of the existing units can be set aside as the spare. We explained why precision air conditioning is necessary and the selection criteria in our data center cooling article.
At YOİ Mühendislik, we carry out load measurement, capacity calculation and single point of failure analysis in server rooms at the site survey, and present our N+1 recommendation in writing. For installation, sales and installation is available, and for existing units our precision air conditioner service; the cost is determined at the site survey.
Frequently asked questions
What is the difference between N+1 and 2N?
In N+1, one spare unit is added to the units that carry the load. In 2N, the entire group that carries the load is installed a second time; it provides higher security but the investment and space requirement increase noticeably.
Does having two air conditioners in the room count as N+1?
It counts only if a single unit can carry the room's entire load. If the load cannot be met without both units running together, the system is not redundant.
Why is the standby air conditioner run in rotation?
Rotation ensures that all units reach similar operating hours and reduces the risk that a standby unit that has been idle for a long time turns out to be faulty when it is needed.
How often should the redundant system be tested?
At least once a year, preferably before summer, one unit should be taken out of service in a controlled way and it should be verified that the remaining units carry the load.