Actually, there is a solution for that... It's called ODAP and it allows your LNS to pull address pools from a server. So you can have smaller pools (like /25's or /24's) assigned from the server and announced as aggregates. Even a /25 is better than 128x/32's
http://www.cisco.com/c/en/us/td/docs/ios/ios_xe/ipaddr/configuration/guide/xe_3s/iad_xe_3s_book/iad_dhcp_sod_apm_xe.html It has been a while since I played with it, but the concept should be mostly the same. Arie -----Original Message----- From: Youssef Bengelloun-Zahr [mailto:[email protected]] Sent: Monday, August 18, 2014 10:17 To: Arie Vayner (avayner) Cc: Mike; [email protected] Subject: Re: [c-nsp] LNS question asr 1002 Hello Arie, I hear you and your arguments are perfectly understandable. The only downside I see with per-LNS pool is lack of redundancy in case of hardware failure. In previous companies I worked for, PPPoL2TP used to terminate randomly on a pool of LNS based on a radius Round Robin algorithm. Excellent for balancing sessions evenly (or not) but the one downside is that you have to re-announce /32s inside your BGP domain. If you RRs can handle it, then why not do it... I guess that this isn't a problem for small to medium sized ISPs, but that's a different song for big ones. Again, it'll all depends on your business case and pre-requisits. Best regards. > Le 18 août 2014 à 18:50, "Arie Vayner (avayner)" <[email protected]> a écrit : > > You may actually want to look at summarizing this. The best practice would be > to have a per-LNS pool (either locally managed or from RADIUS) and advertise > the summary from the LNS up to the network. > You may need to redistribute also connected routes for "fixed IP" services > where a user may have a custom IP from the RADIUS. > > Not summarizing means that every connection (and disconnection) is a BGP > update driving your CPU utilization across the BGP domain... > > > Arie > > -----Original Message----- > From: cisco-nsp [mailto:[email protected]] On Behalf > Of Mike > Sent: Monday, August 18, 2014 09:23 > To: [email protected] > Subject: Re: [c-nsp] LNS question asr 1002 > > >> On 08/17/2014 08:24 PM, Edwardo Garcia wrote: >> Secondly, how does one handle running two LNS servers? How does the >> border router know which edge (LNS) to forward too for a particular >> IP? > > I do it with iBGP where my router is advertising individual /32's. > Yes it makes the route tables longer but it works well in my environment. > YMMV. > > Mike- > > _______________________________________________ > cisco-nsp mailing list [email protected] > https://puck.nether.net/mailman/listinfo/cisco-nsp > archive at http://puck.nether.net/pipermail/cisco-nsp/ > > _______________________________________________ > cisco-nsp mailing list [email protected] > https://puck.nether.net/mailman/listinfo/cisco-nsp > archive at http://puck.nether.net/pipermail/cisco-nsp/ _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
