2008.02.20 NANOG 42 IPv6 hour feedback and closing notes

All done--thanks to CRG West for a kick-ass NANOG 42,
and we'll see you all in Brooklyn in June!


Randy Bush, results of IPv6 hour from yesterday.

Bill Fenner, Duane Wessels, and ???

Larry made it work to the extent that it did.

Grabbed packet headers for the hour; did a
simple distribution of IPv4, IPv6, and ARP.
Should have subtracted out the NAT-PT traffic
from v4, as it gets counted as both.
most of the v6 traffic was NAT-PT; that's why
v4 tracks v6 identically for the hour.

When experiment started, amount of ARP traffic
Duane was pinging the v4 addresses on the subnet;
there were ARPs for offnet addresses.
Some implementation seem to be making on-net
assumption; with no default route, just try
arping for it?
Keep in mind the v6-xp had v4 but no default.

MAC addresses associated to given SSID

blue and white were V6 SSIDs, you can see
that people *did* shift from the v4 to v6

nanog and nanog-a got double-counted, as they
went to the same back-end; but it's still very
clear most people tried switching to v6 SSIDs.

Duane shows the pinging of v4 addresses during
the hour.

People definitely stopped trying to use v4 while
it was turned off, and started again when they
got back from lunch.

Linux NAT-PT
one known existing implementation

Userpace based daemon

had planned on using Cisco NAT-PT...but unable to
make it work.

Moved NANOG stats collector Linux box to trunk port
configured 3 vlans for nanog-v6, nanog-xp-v6,

It puts interface in promiscuous mode, listens
for all packets, does translation in userspace.

Uses DNS lookups, it synthesizes quad-A responses
when there's no A response forthcoming.

Daemon repeatedly appeared to hang
Had to restart it numerous times
Seemed to work better during v6 only hour.

There's also v6 on regular networks, could be
hanging due to nat-pt box seeing v6 on the
main nanog/nanog-a networks (outside).

cacti graphs, total traffic through CRG-West
20m to noon, drops to 2-3M, 12:20 NAT-PT
started working better, hit 10Mb, drops at 1300
for lunch.

total number of associations on APs;
peaked at 314 early, at noon drops down to about 100
clients at noon; went up to about 175 during the hour.

that's just layer 2 association; but the previous
traffic was indeed showing there was some v6 happening.

Collected *lots* of "these need to be fixed" data
as well for vendors.

Jay Hennigan, Impulse Internet--used nanog-v6 and
it's getting much better performance than it was
during the v6 hour.
competition for resource, or need for it to listen
to all three networks, could be a bit of a
reconfiguration that helped get things going.

Closing comments from Todd

Very impressive turnout for a four day NANOG!

Thanks to CRG-West, wonderful hotel, party last night,
many thanks to break sponsors, they help keep the
conference price reasonable for us.

575 registered attendees (483 paid)
20% ISP
21% network operator
18% equipment vendor
12% content provider
6% research
2% government

bunch of ISP folks moving to content houses

top countries
488 US
19 Japan
13 UK
11 Canada
20 others

41 Cisco
16 juniper
14 equinix
13 CRG West

Much related to close travel for them. :wink:

FILL OUT YOUR SURVEY! www.nanog.org!

Need to know what was good, what was bad.

NANOG43 in Brooklyn, June, TelX is sponsoring
it, will be a fantastic, impressive logistical
feat given hotel prices in NY; getting price
under $200/night is very, very impressive.

Program Committee--huge kudos for getting program
out soon enough for people to make plans around

If you have ideas for presentations, come talk to
the yellow badged people! If you can help, talk
to the yellow badge people!