bgp ipv4 filtering

Hi all,

I have to rethink our bgp route filters, as the ipv4 route count is still increasing and now will reach our old devices limits soon.

There is no possibility of upgrade, and a partial bgp table is better to our multi-homed setup that no table and just default routes IMHO.

Last time I added default routes to all our bgp peers and filtered out routes with an as-path length of 20 or longer, routes with a prefix longer than 25, then allowing routes with as-path up to 4 different as (no matter how many prepends). That reduced the installed routes to 715k.

That filter used the philosophy of "near is more important than far", as I thought that to long distances carriers might share infrastructure mostly.

But after having issues from emea to south east asia with some providers, but others not I want to inverse the idea.

I will remove routes with an as-path longer than 20, prefix length of 25 or longer and as path of 3 different as or less. That will reduce the routes to about 445k, at least on my tests.

What do you think?, any ideas would be appreciated.