Current IPv6 deployment methods in datacentres.

Mike Jones mike at
Mon Nov 12 07:42:20 CET 2012


I was just wondering what different methods of IPv6 delivery people
are deploying/seeing in their datacentres.

I have servers with several providers who have just assigned a /48 or
/64 to the LAN (some with RAs and autoconfigure, some requiring static
config) then given out addresses to customers on request (with none so
far apparently preventing me grabbing any address, something i'm sure
will change once I stop being their only customer actually using v6!).
So far I haven't had a provider that has routed a /64 to me as the
default option, which I had expected to be a common thing back when
this IPv6 Internet thing was still theoretical. I did however have one
provider by default assign me a /64 to use out of the /32 that was
configured on the link (glad those ND exhaustion attacks haven't
really been happening in the wild so far!).

Are any of you routing a subnet to all of your customers by default?
is this in addition to or instead of assigning a host address? or has
this idea been largely forgotten as IPv6 has left the theoretical
world and entered real datacentres?

Also, how many of you are actively filtering which IPv6 addresses
customers can use, compared to on v4? I've also noticed a couple of my
providers allow source address spoofing on v6 whilst filtering it on
v4 (in some cases it's actually their upstreams filtering for v6,
while they filter their own v4), obviously not expecting anyone to
reply advertising that their network allows customers to do that ;)

- Mike

More information about the ipv6-ops mailing list