On Sunday, December 22, 2013 02:36:01 PM Adam Vitkovsky wrote: > Anyways even without LFA, on an up to date HW the SPF > (full/incremental) is in order of microseconds. > So I'd checkout the spf logs to see how long it takes to > complete full SFP for 1500 priority 1 prefixes (i.e. > loopbacks) multiply by 10 to account for the metro > networks and if that's still acceptable I'd keep it all > L2. I guess programing the line-cards is still the > biggest holdup and this is solved by the (r)LFA.
I suppose the issue with fast covergence is not how quickly each individual router reacts to the fault, but how long it takes the entire network to do so (the sum of the parts). LFA is helpful because the repair is localized (it's a device problem, not a network problem), and immune to the time taken/needed to run a full SPF after link failure detection, i.e., LFA negates the need to wait for convergence, from the point of view of a single device, and the network at large. Mark.
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
