> Do you use max-path eigbp to ignore the IGP metric to the next-hop/egress PE and to loadshare across these paths? Or do you actually have the requirement to load-share across a direct/ebgp and a remote PE path?
Well it's both actually, in order to take care of all scenarios upfront. It's mostly about dual hub sites that would like to actively use both links to two PEs in a single POP (or two POPs for that matter). If all the remote sites are connected to PEs other than those where hub site is connected to we could get away with: maximum-paths ibgp x unequal-cost. But if later some of the chatty sites have to be migrated to the same PE where one of the hubs is connected to -it'll be causing saturation of that particular hub link while there's still capacity on the other. So anyways the only viable solution is maximum-paths eiBGP after all, and hey it also takes care of the igp-unegual-paths to ibgp NHs as well as ebgp multipathing should there be a need for that. I'd like to add PIC to the solution as well to provide fast restore times also for customer sites with active/backup links. Thank you very much for the BRKIPM-2265 adam -----Original Message----- From: Oliver Boehmer (oboehmer) [mailto:[email protected]] Sent: Wednesday, January 09, 2013 5:38 PM To: Adam Vitkovsky; [email protected] Subject: Re: [c-nsp] IOS XR PIC (+ multipath) > >> assuming you are referring to multipath-protect >I'm sorry I didn't make myself clear -now that I understand -I don't >really need the multipath-protect feature. >What I was actually discouraged to use because of the lack of >real-world deployment was the: backup-path + multipath features - i.e. >using both >commands: "additional-paths selection route-policy" + "maximum-paths >eibgp" >in particular. >I'm interested in the above combination because I'll give us the best >of both worlds, load-sharing for equal cost paths thus implicit backup >as well as backup path for prefixes with single best path. Do you use max-path eigbp to ignore the IGP metric to the next-hop/egress PE and to loadshare across these paths? Or do you actually have the requirement to load-share across a direct/ebgp and a remote PE path? If the latter, I would give it some extra testing, but it should work. > >>I would consider best-external a safe and required feature to achieve >>fast >convergence in active/standby scenarios. >So do I, I guess the SE folks are just scared of "new" things well, quite often us post-sales/Services folks are more cautious ;-) > >Yeah the Cisco Live, >We can't attend this year unfortunately. Any chance your presentation >is going to be one of the online webex sessions? don't think so, but last year's version was videotaped, you should find it (BRKIPM-2265) on www.CiscoLive365.com. Not much changes to this year's preso.. oli _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
