Network redundancy means building a second path or a second device wherever one failure would take down something you cannot do without. In a building, that usually means paired core switches, two uplinks from each telecom room, fibre on two different routes, two internet connections and backup power for the network and its cooling. The skill is in deciding where redundancy is worth it, and then proving it works.
Start with what has to stay up
Redundancy costs money: more switches, more fibre, more ports, more configuration. Before choosing any of it, list what depends on the network and what happens when it stops. Phones, point-of-sale and email are inconvenient to lose. Access control, video recording, nurse call, building automation and remote monitoring of critical equipment may be far more serious.
For each system, ask two questions: how long can it be down before it matters, and does it have its own fallback? A door controller that keeps working offline with cached credentials is less dependent on the network than a video recorder that loses footage while its cameras are unreachable. The answers tell you where to spend.
Where single points of failure hide
A single point of failure is any component whose loss takes down service on its own. In buildings, the usual ones are:
- The core switch. If every IDF connects to one core switch, that switch is the whole network.
- IDF uplinks. One fibre pair from each IDF means one damaged patch cord or transceiver isolates a floor.
- The riser. Two fibres in the same sleeve are not really two paths. A fire, flood or careless trade can damage both.
- The internet connection. One carrier, one entrance, one router.
- Power. Switches with a single power supply on a single circuit, or a UPS with no generator behind it for long outages.
- Cooling. Telecom rooms that overheat when the base building cooling shuts down at night or during an outage.
- Configuration. A redundant pair that was never configured to fail over, or a change pushed to both devices at once.
Dual uplinks and paired cores
The most common redundancy in a building network is a pair of core switches with each access switch connected to both. Under normal conditions both links carry traffic, or one carries it and the other stands by. When one link or one core switch fails, traffic moves to the other path, usually without anyone noticing.
How the failover works depends on the equipment. Some designs bundle the two links into one logical link across a pair of core switches that act as one. Others use routing protocols or spanning tree to choose the active path. Either way, the configuration has to be right, and it has to be tested. A redundant link that has never been unplugged on purpose is an assumption, not a fact.
Diverse physical routes
Redundant links only help if they do not share the same weak point. Two fibres in the same riser sleeve, two carriers whose cables enter through the same conduit, or two core switches in the same rack on the same circuit can all fail together. Physical diversity means the second path takes a genuinely different route through the building.
In existing buildings, a second riser is not always available. Options include routing the second fibre through a different shaft or ceiling space, or at least separating the two cables within the same riser. Whatever is done, record the actual routes on the drawings, because the next contractor needs to know the second fibre is there for a reason.
Power and cooling
Network equipment is only as available as its power. A common arrangement is a UPS in each telecom room to ride through short interruptions and allow an orderly transfer, with the UPS fed from a circuit on the building's standby generator where there is one. Switches with two power supplies can be fed from two separate circuits. PoE switches deserve particular attention, because they power the cameras, access points, phones and card readers plugged into them: if the switch goes down, so does everything it powers.
Cooling is the part most often missed. A small IDF with several PoE switches produces real heat. If its cooling stops while the switches carry on running on UPS and generator power, the room can overheat and the equipment shut itself down anyway. For more on the power side, see data centre power and UPS.
Choosing what to make redundant
Not every building needs every option. The table sets out the common choices and what each protects against, so they can be matched to the systems that depend on the network.
| Option | Protects against | Worth considering when |
|---|---|---|
| Paired core switches | Loss of the core, and downtime during core upgrades | More than a few IDFs, or any critical system on the network |
| Dual uplinks per IDF | A failed fibre, transceiver or port isolating a floor | The IDF serves security, clinical or building systems |
| Second riser route | Physical damage in one shaft or sleeve | Dual uplinks are installed; otherwise they share a weak point |
| Two internet carriers | Carrier outages and cut entrance cables | Cloud services, hosted phones or remote monitoring are essential |
| UPS and generator circuit | Power interruptions and longer outages | Almost always for the core; for IDFs with PoE-powered security devices |
| Spares on site | Waiting for parts after a hardware failure | Always; a cheap complement to every other option |
Testing failover
Redundancy that has not been tested should be assumed not to work. A failover test is planned with the owner, done at a quiet time with someone ready to restore service, and recorded. A typical test sequence:
- Confirm both paths are up and note which is active.
- Disconnect one uplink and record how long service took to recover and whether any device lost connection.
- Restore the link and confirm traffic returns to normal.
- Repeat for the other uplink, each core switch, each internet connection and each power feed.
- Record the results, any problems found and what was changed to fix them.
Repeat the tests after major changes and on a regular schedule agreed with your IT team, since firmware updates and configuration changes can quietly break failover.
How ehZee can help
Our network design service identifies the single points of failure in an existing network or a new design, recommends where redundancy is worth the cost, and plans failover testing as part of commissioning. Power and backup arrangements are coordinated with the electrical design. For the overall layout, see network topology for buildings.
Common questions
Does every IDF need dual uplinks?
Not always. Where an IDF serves systems that must stay up, such as access control, video recording or clinical equipment, dual uplinks are usually worth it. For a small IDF serving only a few desks, a single uplink with spare fibre and a spare transceiver on site may be enough.
Is a second internet connection worth it?
If the building relies on cloud services, remote monitoring or hosted phones, often yes. The two connections should come from different carriers and, ideally, enter the building by different routes, or they may fail together.
How long should the UPS run?
That depends on whether there is a generator. With a generator, the UPS mainly has to bridge the transfer. Without one, it sets how long the network runs in an outage, and the size follows from the load and the required runtime.