[ppml] Policy Proposal: IPv6 Assignment Guidelines
stephen at sprunk.org
Tue Aug 21 12:00:06 EDT 2007
Thus spake "Keith Medcalf" <kmedcalf at dessus.com>
> And more to the point: If every holder of a IPv4 routing slot
> (whether ISP or, like me, a legacy PI /24 holder) was "granted"
> a /28 and they were all routed in the DFZ, do you believe that
> the IPv6 table would contain any more entries than it currently
> does, if the current IPv4 rules were immediately thereafter
> imposed on future IPv6 and ASN allocations/assignments
> (with changes re size in that all IPv6 are /28).
> Would not the table merely double in size (entry-wise that is)?
As of last Friday, there were 25995 ASes and 228981 routes visible in the
DFZ, or 8.8 routes/AS. Of those ASes, 11060 were origin-only ASes only
announcing a single route; the other 14935 ASes accounted for 218921 routes,
or 14.6 routes/AS. Several ASes announce over a thousand routes each, and
many, many more announce hundreds each.
If we were able to give exactly one address block per AS in v6, which is a
goal of many of us, that means (in the absence of TE) the v6 routing table
would be just over one tenth the size of the v4 table. Of course, TE messes
that up, but the impact can be limited by mechanisms that ensure TE
deaggregates only travel a few ASes away instead of polluting routing tables
worldwide. That may not help the largest ISPs much, but at least it'll keep
the problem mostly local to the folks who are causing it.
> Could this be handled in the DFZ at present? If it can then
> there are no worries at all. You then need to evaluate if such a
> happening would be a best-case or a worst-case outcome ...
> or perhaps something in between that can be worked with.
I don't expect the number of routes in the v6 DFZ to be substantially
smaller than with v4 -- but with fewer routes per player, we can accomodate
> If this could not be handled in the DFZ at present, then there is
> a clear need to 'guide' the migration toward a more acceptable
There are ISPs who claim their routers are on the verge of falling over
already, but not because of DFZ size -- they're carrying gobs of internal
and MPLS VPN routes. The DFZ is just another thing they need to find space
for in their routers, and it's not trivial but nor is it anywhere near
today's limits on its own.
Stephen Sprunk "God does not play dice." --Albert Einstein
CCIE #3723 "God is an inveterate gambler, and He throws the
K5SSS dice at every possible opportunity." --Stephen Hawking
More information about the ARIN-PPML