High Availability and Resilient industry standard OSI Layer 2 options for Network Topology in Automation Controls Networks are: Spanning Tree, REP Ring, Device Level Ring, Parallel Redundancy Protocol, Flex Links and EtherChannel.

Spanning Tree(Typically used in both Distribution and Access layers of the network)

Has to manage an entire layer 2 bridged network domain, therefore can be more difficult to manage and convergence could take up to several seconds, even minutes,  depending on the size, spanning tree mode and organization of the network segment.  It’s big win is cost and it’s default configuration options.  As long as you have devices with spanning tree (typically PVST, RVST, PVST+, MST) protocol support in your switches, you can put together a H.A. network than can meet the needs of many simple commercial applications, but not H.A. reliable for Automation Controls networks.  In Controls networks STP is best suited to manage accidental loops.

REP Ring(For use in Distribution layers of a network)

A REP Ring is simply a ring of switches.  A group of switch ports chained together, much like a linear network, but having a closed connection that is self-supervised, not needing a supervisor node device(like DLR).  REP connected switches can redirect traffic in case of a single network failure. 24 switches in a REP ring with a max 100 device node count in the domain have been successfully tested.  Convergence of 100 to 200milli/seconds is obtainable, so many Automation Control applications are tolerant of this timing, but not suited for CIP Sync applications.   Cost plays in a little more since switches that are REP capable are required.  The Cisco IE and Rockwell Stratix managed switches are industry leaders in this space.   NOTE: additional cost savings on cable infrastructure are obtainable.  Additional Network Management is needed to monitor and support this topology, however, less effort compared to Spanning Tree since most of these segments are smaller and purposeful configuration is needed compared to just hopeful defaults as in unmanaged STP applications.

Parallel Redundancy Protocol or PRP(For use in connecting Automation edge devices in the Cell Zone)

PRP may be considered a fully Redundant solution and not just one of High Availability.  PRP does not need to modify the active network topology like the other HA protocols during convergence since it uses to two independent active networks.  Each frame is replicated from the sending node and transmitted over both active networks, while the receiving node simple processes the first frame and discards the second.  NOTE: this does not use configured VLAN technology to separate the channel paths.   Convergence time is non-existent since both networks are active which obviously makes it the best choice for zero tolerant network failure systems, especially with CIP Sync applications.  Management efforts are similar to REP and Device Level Ring since purposeful configuration is required.  As you might guess, this is the most expensive topology since duplicate switches and cable infrastructure is required.

 EtherChannel(Used in the Distribution network layer)

EtherChannel is a redundant star topology that performs logical link port aggregation with two or more ports between switches or server systems running VMWare.  EtherChannel is considered a H.A.  protocol, however, it’s primary advantage is adding the bandwidth capabilities of the physical ports together to increase the throughput capabilities of the link.  Meaning two, 1 Gig ports used in an EtherChannel configuration will yield a 2 gig virtual connection.   Convergence time due to a failed physical link should be less than 250milli seconds which should be adequate for Non-CIP Sync Automation Controls configurations.  Cost for switches run similar to a REP ring since most Cisco and Rockwell switches that support REP also support EtherChannel.  Additional costs due to doubling the cable infrastructure can be incurred, but if frequent cable failures are chronic in a Mfg plant, then the extra cabling pays for itself over time.   EtherChannel is typically the default topology for I.T. Managed Cisco based Distribution Networks.

Flex Links(Typically used in the Distribution network layer)

Flex Links are two ports(port channel or physical interface) on a switch, where one interface is configured to backup the primary.  It works similar to Spanning Tree where the backup link is offline until needed due to a failure of the primary link, however, it is a configured protocol and not automatic like spanning tree can be.  Only 1 backup link per primary port is permitted.  No other function other than Flex Links is allowed for the backup port.  Convergence time is typically less than 50milliseconds and the cost ratio for switches is similar to EtherChannel. It is a faster option for recovery that could provide High Availability for most Automation Controls systems including most CIP Sync applications, with the exception being Motion.

Device Level Ring DLR (For use in connecting Automation edge devices in the Cell Zone)

Similar to a REP Ring for Switches, DLR is a ring topology for Automation Devices.  It is a single fault tolerant ring that has a convergence time typically less than 5milliseconds after a fault, which makes it acceptable for all Automation communications including CIP Sync and Motion applications.  An automation device that wishes to participate in a DLR needs to have the Embedded Switch Technology or utilize an E-Tap.  These required features along with devices that have ring supervisor responsibilities, make the DLR more expensive than other HA options, but still less than PRP.  DLR also requires a more configuration and management than something like spanning tree.  However, we have set several of these up with great success and minor complications.  Device count should not exceed 50, and less is recommended to maintain the HA convergence of less than 5milliseconds.