Cisco Industrial Ring Redundancy: REP vs MRP vs PRP vs HSR vs DLR
Five protocols cover ring and parallel redundancy on Cisco industrial switches: REP (Cisco-proprietary ring convergence), MRP (the IEC 62439-2 ring standard PROFINET uses), DLR (ODVA’s device-level ring for EtherNet/IP), and the two zero-loss IEC 62439-3 siblings — PRP over two parallel LANs and HSR inside a ring. The first three reconverge a broken ring in milliseconds; PRP and HSR duplicate every frame so a failure loses nothing at all.
Which one you run is rarely a free choice. The fieldbus usually decides it — PROFINET lines run MRP, EtherNet/IP cells run DLR, IEC 61850 substations run PRP or HSR — and the Cisco Catalyst IE platform decides the rest, because the zero-loss protocols are gated by switch hardware, not by license tier.
Industrial Ring Redundancy Protocols Compared
| Protocol | Standard | Topology | Recovery behavior | Ecosystem fit | Cisco IE support |
|---|---|---|---|---|---|
| REP / REP Fast | Cisco proprietary | Ring (segment) | Reconvergence, ms-class; detection-time bound | Cisco-only rings, mixed industrial LANs | IE3100–IE9300 |
| MRP | IEC 62439-2 | Ring | Profile-bound: 500, 200, or 30 ms | PROFINET | IE3x00 family, IE9300 |
| DLR | ODVA (EtherNet/IP) | Ring | Within 3 ms (400 µs beacons) | EtherNet/IP device rings | IE3400/IE3400H |
| PRP | IEC 62439-3 | Two parallel LANs | Zero loss — no reconvergence | IEC 61850 substations, process control | IE3400, IE9300 |
| HSR | IEC 62439-3 | Ring | Zero loss — no reconvergence | IEC 61850, ring-shaped bays | IE9300 (IE3400 release-dependent) |
Compiled from Cisco’s redundancy protocol configuration guides for the Catalyst IE3x00 and IE9300 series. Every recovery figure is conditional — on link-detection times, convergence profiles, and beacon intervals — never a flat number.
Why Not Use Spanning Tree in an Industrial Ring?
Spanning tree recovers too slowly and too unpredictably for control traffic. Cisco’s own PRP documentation frames the problem: reconvergence protocols like RSTP, REP, or MRP restore traffic by reopening a blocked port, and that recovery “can take from a few milliseconds to several seconds” depending on the protocol and topology — classic RSTP sits at the seconds end, which stalls PLC cycles and drops IO connections. REP exists precisely because of that gap; Cisco’s REP documentation describes it as an STP alternative built for faster, predictable convergence in segment topologies. Carrier networks solve the same problem with ERPS (ITU-T G.8032), but on the plant floor the choice comes down to the five protocols in this comparison.
The deeper issue is determinism. A PROFINET or EtherNet/IP line does not just need the ring to heal — it needs a guaranteed worst-case number an automation engineer can validate against the controller’s timeout. That is why MRP and DLR publish fixed recovery profiles, and why substations skip reconvergence entirely with PRP and HSR.
What Is REP, Cisco’s Resilient Ethernet Protocol?
REP (Resilient Ethernet Protocol) is a Cisco-proprietary ring protocol that controls a chain of ports called a segment, blocks one alternate (ALT) port per VLAN to prevent loops, and unblocks it the moment a link in the segment fails. Cisco positions it as the STP alternative for ring-shaped industrial and access networks, with VLAN load balancing across two ALT ports as a bonus. It runs on trunk ports only, allows at most two segment ports per switch, and scales to 8 REP segments (3 REP Fast) on a Catalyst IE9300 — raised to 26 open segments in recent IOS XE releases.
REP Segments, Edge Ports, and Failure Recovery
A segment is bounded by two user-configured edge ports, with standard segment ports in between. In steady state every port forwards except the ALT port; on link failure, the switches flush and unblock the ALT port so traffic reroutes around the break. The real-world caveat sits in link-down detection, and Cisco’s guide is unusually candid about it: fiber Gigabit ports detect a dead link in about 10 ms, but copper Gigabit ports take 350 to 750 ms — which is why REP convergence on copper rings disappoints engineers who read only the protocol’s headline. REP Fast closes that gap with a per-link beacon: frames every 3 ms and a 10 ms reception check detect failure within 5–10 ms even on copper, provided both ends of the link run REP Fast.
REP Configuration Example on a Catalyst IE Switch
A minimal segment on two ring ports looks like this (trunk ports, one edge marked primary):
interface GigabitEthernet1/1
switchport mode trunk
rep segment 10 edge primary
!
interface GigabitEthernet1/2
switchport mode trunk
rep segment 10
Verify the segment with show rep topology and confirm every port reports its role before injecting a test failure. Choose REP when the ring is all-Cisco and no fieldbus dictates otherwise; its proprietary nature is the one hard boundary, since third-party switches cannot join a REP segment.
What Is MRP, the PROFINET Media Redundancy Protocol?
MRP (Media Redundancy Protocol, IEC 62439-2) is the ring standard PROFINET networks use: one node acts as Media Redundancy Manager (MRM), blocking one of its two ring ports and probing the ring with test frames, while every other node is a Media Redundancy Client (MRC) that forwards traffic and reports link changes. When the ring breaks, the MRM unblocks its second port and the ring converges within a fixed profile — deterministic by design, and multi-vendor by standard. On Cisco IE switches the default role is MRA (Media Redundancy Automanager), which votes one manager into existence and drops the rest to clients automatically.
MRP Recovery Profiles by Ring Size
Cisco’s configuration guide supports three convergence profiles — 500 ms, 200 ms, and 30 ms — with up to 50 MRCs per ring, and a standalone IE9300 scales to 12 rings on the 200/500 ms profiles from IOS XE 17.13.1. The profile is a worst-case guarantee to validate against controller timeouts, not a typical value: a healthy ring usually beats it comfortably. Cisco lists no separate feature license for MRP, and STP interoperability is handled inside MRP’s CLI mode without extra configuration.
MRP on Cisco IE Switches
MRP runs across the Catalyst IE3x00 family and the IE9300, as both series’ redundancy guides document, with manager and client roles supported (show mrp ring 1 and show platform mrp summary are the working verification commands, plus show profinet mrp ring 1 on PROFINET deployments). Choose MRP whenever PROFINET devices share the ring or the ring mixes vendors — it is the interoperable option REP cannot be.
What Is DLR, the EtherNet/IP Device Level Ring?
DLR (Device Level Ring) is ODVA’s ring protocol for EtherNet/IP: a ring supervisor sends beacon frames every 400 microseconds (the default interval), detects a break from the beacon flow, and unblocks its second port so the ring recovers — Cisco’s guide states nodes resume communication within 3 milliseconds. Every DLR node needs two Ethernet ports with embedded switch technology, one node must be the supervisor, and Cisco recommends a backup supervisor with a precedence value ready to take over.
DLR on Cisco IE Switches
DLR support is an IE3400 story with an FPGA twist. Catalyst IE3400 Rugged and Heavy Duty switches support DLR from IOS XE 17.11.1 as supervisor, backup supervisor, or beacon node — but the default FPGA profile supports only one ring on the base system. Switching the FPGA to the Redundancy profile enables two rings, and with an IEM-3400 expansion module the platform reaches its maximum of three. Redundant-gateway designs (two switches bridging the ring to the wider network) are supported with DLR enabled on both gateways. Budget the FPGA profile change into commissioning, and check the current release notes before assuming DLR on any other IE platform — the Catalyst IE3400 series is the documented home.
What Is PRP, and How Does It Deliver Zero Packet Loss?
PRP (Parallel Redundancy Protocol, IEC 62439-3) removes recovery time instead of shortening it: a Dually Attached Node (DAN) sends every frame twice, simultaneously, into two fully independent networks — LAN-A and LAN-B — with a Redundancy Control Trailer (RCT) carrying a sequence number. The receiver consumes whichever copy arrives first and discards the duplicate; if one LAN fails, traffic continues over the other with zero interruption, because nothing has to reconverge. The cost is structural: two complete networks, double the switching infrastructure, and single-attached devices joining through a RedBox.
PRP on Cisco IE Switches
Cisco implements PRP on the Catalyst IE3400 (per the IE3x00-family redundancy guide) and the Catalyst IE9300, where the switch operates as the RedBox and maintains node and VDAN tables for the devices behind it. The license detail corrects a common assumption: per Cisco’s guide, Network Essentials is sufficient for PRP — only running TrustSec over PRP requires Network Advantage. The gate on zero-loss redundancy is platform hardware, not license tier. Choose PRP for IEC 61850 substations and process networks where a lost frame is unacceptable and the site can physically cable two independent LANs.
What Is HSR, and Why Do Rings Need RedBoxes?
HSR (High-availability Seamless Redundancy, IEC 62439-3) achieves PRP’s zero-loss result inside a single ring: each Doubly Attached Node (DANH) sends every frame in both directions around the ring, an HSR header sequence number identifies the duplicate, and the first copy to arrive wins. No reconvergence, no lost frames — and no second network to build, which is HSR’s economic case over PRP. The trade is bandwidth and membership: every frame travels twice around the ring, nodes forward constantly, and only HSR-capable devices can sit in the ring itself.
That last constraint is why RedBoxes matter. Standard, single-attached devices (SANs) join through a RedBox, which duplicates and discards on their behalf and presents them to the ring as Virtual DANs (VDANs), generating their supervision frames. On Cisco’s implementation the Catalyst IE9300 acts as that RedBox in HSR-SAN mode, with two ring-facing Gigabit ports and the loop-avoidance rule that a RedBox never retransmits a frame already sent in the same direction.
HSR on Cisco IE Switches
The IE9300 supports HSR-SAN in standalone deployments with two hard limits from the configuration guide: one HSR instance per switch, and HSR and PRP are mutually exclusive — a switch with a PRP channel cannot also host an HSR ring. Network Essentials covers the feature, and HSR stays disabled until explicitly configured. IE3400 support is release-dependent, so verify the IE3400 redundancy guide for your IOS XE version before designing around it.
What Is the Difference Between PRP and HSR?
PRP and HSR deliver the same guarantee — zero packet loss on a single failure — through different geometry: PRP duplicates frames across two independent parallel networks, while HSR duplicates them in both directions around one ring. That single difference drives every practical trade-off between them.
| Dimension | PRP | HSR |
|---|---|---|
| Topology | Two independent LANs (A/B) | Single ring |
| Infrastructure cost | Double network build | One ring, no second LAN |
| Bandwidth cost | Duplicates cross two separate LANs | Every frame circles the ring twice |
| Standard devices | Attach to either LAN, or via RedBox | Must enter through a RedBox (VDAN) |
| Duplicate handling | RCT trailer + sequence number | HSR header + sequence number |
| Typical deployment | Substation bus, concentrated equipment | Bay-level loops, distributed equipment |
Choose PRP when equipment concentrates around existing LAN infrastructure or standard Ethernet devices must attach freely; choose HSR when the physical layout is already a loop and a second network is not fundable. Hybrid designs couple both — an HSR ring hanging off a PRP backbone — which is exactly why the mutual-exclusion rule on a single Cisco RedBox matters at design time.
PRP vs MRP for Recovery Time and Cost
PRP buys zero loss with a second network; MRP buys a bounded 30–500 ms recovery with a single ring — the decision is whether the process tolerates a profile’s worth of interruption. A PROFINET line whose controllers ride out 200 ms has no reason to fund LAN-B; a protection relay circuit that trips on one lost GOOSE frame has no business waiting for any reconvergence at all. Cost follows the same line: MRP reuses the ring you already have and needs no feature license on Cisco IE switches, while PRP doubles switching hardware before the first frame flows. The two rarely compete head-to-head — when they seem to, the real question is usually whether the requirement is availability (MRP suffices) or continuity (PRP is mandatory).
PRP vs DLR on EtherNet/IP Networks
DLR is the native EtherNet/IP answer and recovers within 3 ms; PRP appears in EtherNet/IP plants only where even 3 ms is too much or where the design standardizes on IEC 62439-3 across cells. DLR keeps the drive-level economics — one ring, embedded two-port devices, a supervisor in the IE3400 — while PRP demands dual NICs or RedBoxes for every participating device plus the second LAN. In practice the split is hierarchical: DLR rings at the machine and cell level, PRP (or HSR) higher up where substation-grade continuity requirements creep into the process network. Sizing a cell ring on PRP “to be safe” typically doubles the network bill to solve a problem the 3 ms ring never had.
How Fast Does Each Ring Protocol Actually Recover?
Recovery numbers only mean something with their conditions attached — the same REP ring recovers an order of magnitude slower on copper than on fiber, and MRP’s figure is a profile ceiling, not a measurement.
| Protocol | Published behavior | The condition that governs it |
|---|---|---|
| REP | Reconvergence after detection | Link-down detection: ~10 ms on GE fiber, 350–750 ms on GE copper |
| REP Fast | Detection within 5–10 ms | 3 ms beacons; both link ends must run REP Fast |
| MRP | 500 / 200 / 30 ms profiles | Worst-case guarantee; up to 50 MRCs per ring |
| DLR | Within 3 ms | 400 µs beacon interval; supervisor-based detection |
| PRP | Zero loss | No reconvergence — duplicate frames on LAN-A/LAN-B |
| HSR | Zero loss | No reconvergence — duplicates both directions on the ring |
Figures from Cisco’s IE3x00 and IE9300 redundancy protocol configuration guides. Non-DLR devices in a DLR ring worsen its worst case, and REP on GE copper without REP Fast is the most common source of “the datasheet lied” complaints — the protocol was waiting on the PHY.
Which Cisco IE Switch Supports Each Protocol?
REP and MRP run broadly across the industrial lineup, while the zero-loss and DLR features concentrate on the platforms with the hardware to run them — and license tier is not the differentiator.
| Protocol | IE3100 / IE3200 / IE3300 | IE3400 / IE3400H | IE9300 | License note |
|---|---|---|---|---|
| REP / REP Fast | ✓ | ✓ | ✓ (8 segments / 3 fast; 26 open segments on recent releases) | — |
| MRP (MRM/MRC/MRA) | ✓ | ✓ | ✓ (12 rings, 17.13.1+) | None required per Cisco’s guide |
| DLR | — | ✓ (17.11.1+, FPGA profile sets 1–3 rings) | Confirm current release notes | — |
| PRP | — | ✓ | ✓ (RedBox, node/VDAN tables) | Network Essentials; TrustSec over PRP needs Advantage |
| HSR (HSR-SAN) | — | Release-dependent | ✓ (single instance, exclusive with PRP) | Network Essentials |
Per the Catalyst IE3x00 and IE9300 redundancy protocol configuration guides; confirm your exact model and IOS XE release before ordering. For a deeper look at the two platforms that carry the advanced protocols, the IE3300 vs IE3400 comparison covers what the IE3400’s hardware adds beyond redundancy.
The pattern to plan around: a Catalyst IE3300 covers REP and MRP rings at access scale, the IE3400 adds DLR and PRP for EtherNet/IP cells and substation edges, and the IE9300 carries the full set including HSR RedBox duty at aggregation.
How to Choose the Right Ring Redundancy Protocol
Start with the fieldbus, then the loss tolerance, then the topology you can afford:
- PROFINET on the ring? → MRP, sized to the 30/200/500 ms profile your controllers tolerate.
- EtherNet/IP device ring? → DLR on IE3400-class hardware; plan the FPGA Redundancy profile if you need more than one ring.
- Zero frame loss required (IEC 61850, protection, motion)? → PRP if you can build two LANs or equipment concentrates near existing infrastructure; HSR if the layout is a loop and a second network is not fundable. Remember one Cisco RedBox cannot run both.
- All-Cisco ring, no fieldbus mandate? → REP — and REP Fast on any Gigabit copper links, or the PHY’s 350–750 ms detection will eat the benefit.
- Mixed vendors, no fieldbus mandate? → MRP, the standards-based option third-party switches can join.
The ecosystem picks the protocol far more often than a benchmark does; the engineering work is validating the recovery behavior against the controller’s timeout and buying the platform tier that actually runs it.
Troubleshooting Ring Convergence on Cisco IE Switches
Slow or failed ring recovery almost always traces to one of four causes, and each has a fast check:
| Symptom | Likely cause | First check |
|---|---|---|
| REP ring converges in hundreds of ms | GE copper link-down detection (350–750 ms), REP Fast absent | Confirm REP Fast on both ends of every copper link |
| REP segment never completes | Port not trunking, or >2 segment ports on one switch | show rep topology for role and state per port |
| MRP ring stuck open or flapping | Two managers, or profile mismatch across nodes | show mrp ring 1 and show platform mrp summary |
| DLR ring won’t form a second ring | FPGA still on default profile | Verify the Redundancy FPGA profile on the IE3400 |
| PRP shows duplicate discards climbing on one LAN only | LAN-A/LAN-B asymmetry or a failed path nobody noticed | Compare per-LAN statistics in the PRP show commands |
Zero-loss protocols invert the troubleshooting logic: a PRP network with a dead LAN-B keeps working silently, so the failure you must catch is the loss of redundancy, not the loss of traffic. Alarm on per-LAN counters and supervision frames, or the second failure becomes the first outage.
Frequently Asked Questions
Can REP interoperate with non-Cisco switches?
No. REP is Cisco-proprietary, and a segment must be built entirely from REP-capable Cisco ports. For a multi-vendor ring, MRP is the standards-based choice (IEC 62439-2), which is exactly the role it plays in PROFINET networks.
Can PRP and HSR run on the same switch?
No. Cisco’s IE9300 configuration guide is explicit: the switch supports one HSR or one PRP instance, and once a PRP channel exists an HSR ring cannot be created on the same switch. Coupled PRP-HSR designs therefore split the roles across separate RedBoxes.
How many nodes can an MRP ring have?
Up to 50 clients (MRCs) per ring on Cisco IE switches, with the 30, 200, or 500 ms profile setting the worst-case recovery. A standalone IE9300 scales to 12 rings on the 200/500 ms profiles from IOS XE 17.13.1.
Do PRP or HSR require a Network Advantage license?
No — Cisco’s IE9300 guide states Network Essentials is sufficient for both, with one exception: running TrustSec over PRP requires Network Advantage. Platform hardware, not license tier, is what gates the zero-loss protocols to the IE3400 and IE9300.
Does DLR need a special FPGA profile on the IE3400?
For more than one ring, yes. The default FPGA profile supports a single DLR ring on the base system; the Redundancy profile raises it to two, and adding an IEM-3400 expansion module reaches the platform’s maximum of three rings.
References
- Cisco, Redundancy Protocol Configuration Guide, Catalyst IE9300 Rugged Series — REP, MRP, PRP, and HSR chapters: cisco.com
- Cisco, Redundancy Protocol Configuration Guide, Catalyst IE3x00, IE3400 Heavy Duty, and ESS3300 Series — Device Level Ring chapter: cisco.com
- IEC 62439-2 (MRP) and IEC 62439-3 (PRP/HSR), International Electrotechnical Commission: webstore.iec.ch
- ODVA, EtherNet/IP and Device Level Ring specifications: odva.org
