[ppml] too many variables

Paul Vixie paul at vix.com
Thu Aug 9 13:12:12 EDT 2007

> From: bmanning at vacation.karoshi.com
> To: ppml at arin.net, nanog at nanog.org

yikes.  cross posting between two public lists seems like a really bad idea.
i'm going to follow this up on ppml only.

>  I asked this question to a couple of folks:
> 	"at the current churn rate/ration, at what size doe the FIB need to 
>          be before it will not converge?"
>  and got these answers:

none of which were germane.  given infinite cpu and infinite RAM and a stated
DAG size and a stated churn rate and a stated average interprovider link size
and a stated headroom target on those links, at what point is there not enough
bandwidth to carry the updates?  then, if bandwidth isn't a consideration but
the speed of light remains as stated, at what point does convergence stop
being theoretically possible?  once we know those limits, we can ask questions
to which the last set of answers were germane, like how many megacomputrons do
we need in core routers, edge routers, tier 1 routers, tier 2 routers, etc.
and from that, finally, we could guess reasonable maximum RIB and FIB sizes
for a worst case definition of "can reach everything from".

More information about the ARIN-PPML mailing list