:-) I tried this already, but the match metric only works with IGRP route-maps, not BGP :-) -Andy -- Andy McConnell アンディ マッコネル Network Architect, NTT Multimedia Communications Laboratories Calvin: People think it must be fun to be a super genius, but they don't realize how hard it is to put up with all the idiots in the world. Hobbes: Isn't your pants' zipper supposed to be in the front? On 10 Mar 1998, Richard Parker wrote: richard_parker> Andy McConnell wrote: richard_parker> >Avi Freedman suggested using a +1 metric when leaving each member-AS. richard_parker> >But it doesn't seem to help - perhaps I didn't do it right. richard_parker> >In fact, it doesn't look like the metrics are adjusted more than 1. richard_parker> >... richard_parker> >Setting the MED is the crux of the problem - I can set it "statically", richard_parker> >but the idea is to make the whole confederation dynamic. I want the richard_parker> >confederation to work like a true "internet". (Yes, I'm asking for the richard_parker> >world...) richard_parker> richard_parker> Andy, richard_parker> richard_parker> Perhaps a solution, albeit an ugly solution, to the problem is to richard_parker> use a large route-map that matches and increments the MED for the richard_parker> expected range of MED values. Each router would have a route-map richard_parker> to increment the metric on updates that it sends. On the Cisco richard_parker> this configuration might look something like the following: richard_parker> richard_parker> neighbor a.b.c.d route-map INC-MED out richard_parker> ! richard_parker> route-map INC-MED permit 1 richard_parker> match metric 0 richard_parker> set metric 1 richard_parker> route-map INC-MED permit 2 richard_parker> match metric 1 richard_parker> set metric 2 richard_parker> (etc.) richard_parker> richard_parker> If dorian's remark is true... richard_parker> richard_parker> >This won't work because Cisco's bgp implementation skips over MED richard_parker> >from neighboring sub ASes inside a confederation in its decision richard_parker> >process. richard_parker> richard_parker> ...then instead use a route map to increment the local preference richard_parker> for each route received. You would still probably want to set the richard_parker> MED from the local preference before you export routes so that the richard_parker> external neighbors (in other ASes) of the confederation members richard_parker> choose to send data into the confederation at the point closest to richard_parker> the ultimate destination. richard_parker> richard_parker> -Richard richard_parker>