IPv6 teredo blackout June 8th

Phil philip_mm at me.com
Thu Feb 3 10:13:29 CET 2011

Thanks guys, great response.

I don't think it will not be a potential issue and further testing will be
done on my end will let you know how it goes.
large networks, even with an increase of 5% traffic swapping into the
relays, would see huge throughput issues.
Bernhard brought up a good point concerning the amount of applications
(peer-to-peer, the initial handshake going over DNS, DNS initiated
streaming preferring v6), browser plugins & legacy browsers, etc ... that
still force teredo in preference.  Not to mention standard native IPv6
connectivity that initiates teredo.  I am just curious of what we have not
considered really.

Most windows platforms do have teredo on a default but un-initiated and
requires the application, browser, etc ... to bring it to life, although
W7 requires pre-configuration before its use.

But if it all, or partially, breaks at least it might end up being an
educated exercise and I don't think taking down, filtering and
non-advertisement, of our relay addresses is the answer.
We have 9 months to get it right before the local registries start to run
out and we need a few learning experiences beforehand.  And just on that
note I doubt if we will last that long.  I am expecting June, early July,
for the first registry to run out as the standard mathematically predicts
have taught us on their over confidence previous these dates tend to move
backwards rather extremely, unless RIPE etc policies tend on the deny more
and then we have really run out right now if 4 out 5 get denied on each

6to4 static tunnels may also be an issue but for me not such a priority at
present as there tend to be controlled by the little more savvy, auto
tunneling is another subject in itself.

Our estimates are 16-24 hours for everything to return to normal with a
little cache flushing if the outage becomes permanent enough for a
rollback on the day.
So in addition the load over native v6 itself may also, as mentioned,
cause throughput issues with certain NPUs depending on the feature.
And there are some L2s out there that fall over on 20-30 meg throughput of

I think the result will be interesting, but I think we should rename it to
IPv6 World Test day :)

On 02/02/2011 21:16, "Martin Millnert" <martin at millnert.se> wrote:

>as others have pointed out, content providers dual-stacking their
>frontend web servers is a non-issue for Teredo relays in terms of change
>of traffic load. Will have no impact.
>What can impact the load on Teredo relays is a change in behavior of
>people using any form of native or tunneled IPv6 connectivity, primarily
>in IP address-based host-to-host connections, and of course a change in
>the amount of operating Teredo relays itself.
>On Wed, 2011-02-02 at 11:05 +0100, Phil wrote:
>> Guys,
>> I want to open up a discussion on the potential failure of the teredo
>> relays around the world on IPv6 day and anything we can do to mitigate
>> it.
>> June 8th IPv6 world day all major services come online, and within the
>> network I architect that would increase my traffic, in estimate, to
>> 40-60 gig.
>> Our traffic flows through two relay servers in switzerland at present
>> that have gone down 3 times in the last 14 months.
>> Some windows platforms have teredo enabled by default.
>> AAAA is preferred over A.
>> That would indicate all IPv6 traffic would be preferred through the
>> teredo tunnel with obvious capacity and relay throughput issues
>> suddenly with no way of controlling the customer subs once the DNS
>> entry comes up.
>> Thoughts and mitigation?
>> Cheers,
>> Phil

More information about the ipv6-ops mailing list