On 2008-05-12, Peter Bristow <pete.bristow@gmail.com> wrote:
quoted text
> Thanks for the replies so far. bgpd can handle full tables. This means that
> as of today ~251K prefixes demonstrably work but most of these are from
> peers (iBGP or eBGP) I'm interested in known limitations/expected gotchas on
> locally originating 50k prefixes.50K lines of
>
> network 92.48.111.0/26 set { nexthop 92.48.95.196 community 64667:0 }

I think you should setup a test machine, generate a config and try
parsing it (and also reloading it), looking at how much memory it uses.
15k lines of filter rules (allow...set) was pretty tedious and chewed
through RAM a bit, I don't know if this translates directly to
"network...set" but if it does, you may bump into maximum process
size limits on most arch (1GB on i386).

If it does work out ok though, 29550 would be a nice addition to
http://www.openbgpd.org/users.html ;-)