DLR or PRP: Choosing Redundancy for an EtherNet/IP Network

How DLR works

Device Level Ring (DLR) is an EtherNet/IP protocol defined by ODVA to detect and recover from a single point of failure on a ring network (ENET-AT007E p. 7). Each ring has three roles:

• Ring supervisor — manages the ring. In normal operation the supervisor blocks one of its two ports so the ring does not form a loop, but still sends beacons out of both ports. When beacons stop travelling all the way round, the supervisor unblocks the port and traffic flows in both directions to either side of the break (p. 7–8, p. 12).
• Ring participant — a device in the ring: forwards data and reports fault locations to the supervisor.
• Redundant gateway (optional) — a switch connecting the ring to the rest of the plant network.

The supervisor function is disabled by default on supervisor-capable devices, and every ring must have at least one supervisor (p. 7). A ring without a supervisor is an uncontrolled loop that can cause a broadcast storm and bring down the whole network (p. 8).

Numbers to remember

Rings above 50 nodes must be tested before going into production: the more nodes, the higher the chance of multiple simultaneous faults and the longer the detection and recovery times (p. 9).

Rules for building a ring

1. Only one active supervisor. The device with the highest precedence wins (on a tie, the higher MAC address wins); other supervisors become back-ups. Rockwell recommends configuring at least one back-up supervisor (p. 8–9).
2. Only DLR-capable devices connect directly to the ring. Single-port or non-DLR devices connect through a DLR-capable Stratix switch or a 1783-ETAP EtherNet/IP tap (p. 9).
3. One speed per ring — either 100 Mbps or 1 Gbps. Mixing 1756-EN2TR (100 Mbps) and 1756-EN4TR (1 Gbps) is possible, but port speeds must be set as instructed: leave auto-negotiate, except that the link between two adjacent 1 Gbps nodes in a 100 Mbps ring must be fixed at 100 Mbps (p. 10–11).
4. DLR ports are not configured with STP, REP, EtherChannel, Flex Links, PRP, NAT, 802.1x or MAC-based port security; DLR ports must be full duplex (p. 10).
5. Order for closing the ring: wire it as a line, leaving one segment unconnected; configure the supervisor and DLR on every switch; connect the last segment last (p. 14–15).

Connecting the ring to the plant network

There are three common layouts: a device ring (a tap acting as supervisor), a device ring with Stratix switches (the switch acts as supervisor, collects diagnostics and provides the uplink), and a switch-only ring (p. 17–19). With two redundant gateways, this function is only available on Stratix switches, and every device in the ring must be compatible with redundant gateways, otherwise connections may drop when the two gateways swap roles (p. 22, p. 24). Stratix 5400 and 5800 support up to three rings per switch (p. 29).

A broken ring with the machine still running — it must be monitored

This is the point most easily forgotten: after one break, the I/O keeps running normally, so nobody knows the ring has lost its redundancy until a second break occurs. The application should monitor ring status (p. 12). The supervisor reports the status (0 normal, 1 ring fault, 2 unexpected loop, 3 partial fault, 4 rapid fault/restore cycle) and the IP/MAC addresses of the two nodes on either side of the break (p. 95). This information can be read from Logix Designer, the module's web page, Stratix Device Manager, or a ready-made DLR faceplate for the HMI (p. 89–93).

PRP: two parallel networks, no recovery time

The Parallel Redundancy Protocol (IEC 62439-3) sends every packet simultaneously over two independent networks, LAN A and LAN B; the receiver keeps the first copy and discards the second, so when one network fails there is no recovery time and neither network has to reconfigure (ENET-AT006D p. 7). Devices with two PRP ports are called DANs; ordinary devices connect to both networks through a RedBox switch.

Mandatory rules: never connect LAN A directly to LAN B; the two networks share no switch; a DAN uses the same subnet on both networks; infrastructure switches must allow an MTU of at least 1506 bytes (p. 16, p. 21). On ControlLogix, 1756-EN2TP is the PRP module; 1756-EN4TR supports PRP from firmware 4.001 (1756-TD003R p. 4). Stratix 5400, 5410 and 5800 can act as a RedBox (ENET-AT006D p. 10–15, p. 41). On Stratix 5700, DLR is only available on certain catalog numbers — check the catalog table before selecting (1783-TD002G p. 44). Note: Rockwell discontinued Stratix 5700 on 01/09/2024 and Stratix 5400 on 30/01/2025 (product pages on rockwellautomation.com); the replacement line for Stratix 5700 is Stratix 5200 — 1783-CMS10DN and 1783-CMS20DN offer 2 DLR rings and PRP (1783-TD002G p. 29). The two methods can be combined: a DLR ring behind a RedBox, with every node in the ring protected by PRP (p. 11).

Quick comparison

Table source: ENET-AT006D p. 7.

How to choose

DLR suits distributed I/O, drives and HMIs within one area — less cabling, no extra switches, recovery within a few milliseconds. PRP is for processes that cannot accept any interruption at all and can justify the cost of two networks. Both must be tested by pulling real cables during FAT, not just on paper. DLR-capable catalog numbers include: 1756-EN2TR, 1756-EN4TR, 1734-AENTR, two-port PanelView Plus 7 terminals (catalog numbers containing "22"), two-port CompactLogix 5370, and DLR-capable Stratix switches. Where an existing network runs on ordinary switches, TDL surveys it and redesigns it on managed Stratix switches.