Networking · 8 min read · Updated June 2026
Network Redundancy on the Plant Floor: RSTP, Ring and Beyond
A single broken link should never stop a line. How industrial ring protocols work, when RSTP is enough, which switches to buy, and how to design recovery times that PLCs can live with.
Office networks tolerate seconds of failover. A motion controller does not. Designing plant-floor redundancy means matching the recovery time of the network to the timeout of the protocols running on it — and buying the switches that can deliver it.
Why RSTP alone often is not enough
Rapid Spanning Tree typically reconverges in one to several seconds on industrial topologies. That is acceptable for SCADA polling and monitoring, but a PROFINET device with a 3 × 32 ms watchdog, or an EtherNet/IP connection with a 100 ms RPI multiplier, drops long before RSTP finds the new path. The line stops, then restarts — which is exactly what redundancy was supposed to prevent.
Ring protocols: purpose-built recovery
Industrial vendors implement ring redundancy with deterministic failover:
- MRP (Media Redundancy Protocol, IEC 62439-2) — vendor-neutral, 200 ms guaranteed recovery with up to 50 switches; 30 ms profiles on capable hardware. Native in PROFINET devices, so an ET 200SP station can be a ring client.
- Turbo Ring / Turbo Chain (Moxa), X-Ring, HiPER-Ring and similar proprietary rings — 20–50 ms typical, single-vendor in the ring. The EDS-408A and EDS-G4012 support both Turbo Ring and MRP.
- PRP/HSR (IEC 62439-3) — zero recovery time via parallel paths, used in substations and safety-critical systems at the cost of duplicated hardware.
A ring costs one extra cable run and buys the plant a network that survives any single fibre cut or switch failure.
Design rules that keep you out of trouble
- Close rings through different cable trays and building routes. Redundancy through the same conduit is not redundancy — one excavator removes both paths.
- One ring manager per ring, documented. Two managers, or none, is the most common misconfiguration we find.
- Never mix unmanaged switches into a redundant topology. Unmanaged switches such as the SCALANCE XB008 or FL SWITCH 2008 are excellent inside a panel; they do not belong in the ring.
- Test failover with a real unplug during FAT, and measure PLC and drive behaviour, not just ping.
- Use fibre between buildings and anywhere ground potential differences or lightning are a risk.
Where to spend the money
Put managed, ring-capable switches on the backbone and on machine loops that stop production. Simple star topologies on unmanaged switches remain fine inside a single panel. In a typical plant: a Gigabit managed ring between buildings and cells, Fast Ethernet managed rings within cells, unmanaged switches in machine panels.
| Location | Switch class | Example |
|---|---|---|
| Plant backbone, cameras, high bandwidth | Gigabit managed, PoE, IEC 62443 | Moxa EDS-G4012 |
| Cell ring, SCADA, substations | Fast Ethernet managed, Turbo Ring/MRP | Moxa EDS-408A |
| Inside a machine panel | Unmanaged | SCALANCE XB008, FL SWITCH 2008 |
We design, supply and commission redundant industrial networks — including the MRP/PROFINET validation reports your customers can audit. See engineering services and our guide to OT network segmentation.
Need engineering help with this?
Our engineers apply this to real projects every week — selection, specification, commissioning. Send your question or a part number.
Quick answers
Can I build a ring with unmanaged switches?
No. A loop of unmanaged switches creates a broadcast storm that takes the network down. Ring redundancy requires managed switches running a ring protocol such as MRP or Turbo Ring; unmanaged switches belong in star topologies inside a panel.
MRP or a proprietary ring?
MRP (IEC 62439-2) is vendor-neutral and guaranteed at 200 ms (30 ms on capable hardware), which satisfies most PROFINET and SCADA networks. Proprietary rings such as Moxa Turbo Ring recover faster (under 20 ms) but require switches from one vendor in the ring.