I became aware of the Resilient Ethernet Protocol (REP) a few weeks ago and have been reading about it. The Cisco documents position it as a replacement for the Spanning Tree Protocol (STP). The documentation says that it can fail over in 50ms, which is certainly impressive, but then says that in some cases it can take up to 250ms to fail over without specifying the conditions in which this may occur. If you know any details, please leave a comment.
REP requires a linear topology with the ends of the line connected into the rest of the L2 topology. See Figure 1 below (figures are taken from Cisco docs).
The target use is for Metro Ethernet deployments by service providers, thus the linear topology which is wrapped into a ring, see Figure 2. Each end of the linear topology must connect into the remaining L2 topology so that there is L2 connectivity between the Edge Ports. Figure 2 doesn’t show a blocking port, which is required for normal operation.
You select one port in the path to be the blocking port. In essence, you are creating the spanning tree and identifying the blocking links. When a failure of any other link occurs in the path, the adjacent switches report the failure to their neighbors and the blocking link begins forwarding. Redrawing the networks of Figure 4 using a spanning tree layout is an interesting exercise.
Hosts that are connected to each half of the REP “ring” will normally communicate via the path over the rest of the network, i.e. via the L2 network where the Edge Ports connect. When a failure occurs in the path, the blocking link begins forwarding and the switches all flush their forwarding tables. The forwarding entries are quickly re-learned and the two hosts then communicate via the new forwarding path as shown in Figure 6.
When planning a REP network, it may help to redraw it as a spanning tree topology to determine the best location for blocking ports. Traffic flows may drive your selection of the blocking port as much as balancing the branches of the tree.
Finally, there is a hardware and IOS requirement. It is only available on Metro Ethernet devices like the ME3400. Other devices that are typically used by service providers may support it; the Cat 4500 is one example.
REP is an interesting idea. I think that the trade-off is the manual configuration versus the rapid fail-over time. As service providers work to improve network availability, shaving seconds off the fail-over times will be important and this is one of the tools that is available.
Re-posted with Permission
NetCraftsmen would like to acknowledge Infoblox for their permission to re-post this article which originally appeared in the Applied Infrastructure blog under http://www.infoblox.com/en/communities/blogs.html