What Is Network Redundancy?

Network redundancy is the practice of giving a network more than one way to do each job, so that the failure of one part does not stop communication. The spare part can be a second cable, a second switch, a second gateway router or a second power supply. A redundancy protocol then decides which path or device carries traffic, and how fast the backup takes over when the active one fails.

Redundancy on its own is not enough. Two Ethernet links between the same switches, with nothing managing them, form a loop that floods the network within seconds. A redundant network therefore always pairs extra hardware with logic: spanning tree or a ring protocol to block the spare path until it is needed, link aggregation to share load across parallel cables, or a gateway protocol so hosts keep a working default route.

The useful measure is recovery time, the gap between a failure and traffic flowing again. IEC 62439, the international standard for high-availability industrial networks, compares it with the application’s grace time: the longest interruption the process can tolerate before it trips, stops or raises an alarm.

Redundancy is built in layers, and each layer protects against a different failure. A well-designed plant network usually combines several of them.

The protocol sets the recovery time, so it is chosen to fit the grace time of the process. The figures below are typical values given by the standards or quoted by switch vendors. Real results depend on the number of switches, link speeds and configuration.

The table shows the main trade-off. PRP and HSR send every frame twice, over two paths, so the receiver simply discards the duplicate and nothing is lost when a path fails. That needs dual-port end devices or redundancy boxes and, for PRP, two complete networks. Ring protocols such as ERPS need nothing special from the end devices, and recover in tens of milliseconds, which is fast enough for most control traffic.

In an office, a few seconds of network loss is an annoyance. On a production line, a substation or a train, it can stop a process. Industrial protocols watch their connections closely: PROFINET and EtherNet/IP devices typically drop a cyclic connection after a few missed cycles, often tens of milliseconds, and a dropped connection can put drives and outputs into a safe state. That is why recovery time is specified alongside the cabling.

It is also why ring topologies are so common in plants. A ring of switches uses the least fibre or copper for one spare path, and a dedicated ring protocol recovers far faster than spanning tree. Our explainer on how ring topology works compares the ring protocols, and the page on spanning tree protocol covers how loops are blocked in a mesh.

Our managed industrial switches list ITU-T G.8032 ERPS alongside STP, RSTP and MSTP, with ring recovery quoted at under 50 ms and some models at under 20 ms. Many of our unmanaged industrial switches still take two redundant DC inputs, so the power side of the design is covered even where no protocol is needed.

A redundant network is only as good as its weakest single point. Work through the design in this order.

We supply industrial switches as part of a complete system, and as standard we configure the redundancy protocol, ring ports and power inputs to your network drawing before it ships.