On Apr 17, 2012, at 3:12 PM, Tony Li wrote: > On Apr 17, 2012, at 2:47 PM, Robert Raszuk wrote: > >>> But then the overhead become untenable. >> >> So why not event driven rsync like mechanism between BGP_Adj_Ribs for >> incremental updates ? *well for VPNs RTC would need to be mandatory* > > > That doesn't scale particularly well. The time needed to 'sync' is still > linear with respect to the number of prefixes, and you end up sending a > digest of your database. When the items involved gigabit files, this is an > efficiency win. When the routes are in the hundreds of bytes, you're saving > maybe 2 orders of magnitude over the entire database. > > Sure, that's better, but that's a fixed, one time multiplier on scale.
Exactly right. Just to add, the cost for the digest isn't just transmission, it's also CPU and storage. As you've already pointed out, BGP for all its warts doesn't have this particular issue. --John _______________________________________________ GROW mailing list GROW@ietf.org https://www.ietf.org/mailman/listinfo/grow
