Re: Relative performance of IPv6 and IPv4 for current users

[David Malone <dwmalone@maths.tcd.ie>]

[ post by non-subscriber.  with the massive amount of spam, it is easy to miss
  and therefore delete posts by non-subscribers.  if you wish to regularly
  post from an address that is not subscribed to this mailing list, send a
  message to <listname>-owner@ops.ietf.org and ask to have the alternate
  address added to the list of addresses from which submissions are
  automatically accepted. ]

On Wed, Jul 02, 2003 at 10:23:20AM +0100, Tim Chown wrote:
> > I really wonder if there is a dearth of 6to4 relay routers. My work
> > and home ISPs both have relay routers. When I use 6to4 with other
> > Irish ISPs I see the traffic going to several different relays (I
> > think I've seen two distinct ones on Germany and Switch in Switzerland).
> 
> Many networks have 6to4 relays that they do not advertise outside their
> own network... so "dearth" is a visibility issue also...

If anyone is interested, I had a go at counting 6to4 relay routers
from the IPv6 side. I did this by tracerout6ing with a 6to4 source
address and then using tcpdump to catch the IPv4 addresses of
relay-routers that was used to encapsulate the replies. To decide
where to traceroute to, I used the ::1 addresses of each prefix in
the IPv6 BGP table (as seen from HEAnet this morning).

After doing this, I got replies encapsulated 26 different source
IPv4 addresses. One was the anycast address and I had a go at
guessing the countries of the remainder (AU:1 CZ:1 DE:1 FI:1 IE:1
JP:1 KR:3 LT:2 NL:3 NO:1 PT:1 SE:2 TH:1 TW:1 UK:2 US:3).

About 1/4 of the response packets came from the 6to4 anycast address,
so it seems to be worth trying to figure out how many relays it
represents. Looking at the histogram of TTLs in the IPv4 header,
it looks like there are at least 3 or 4.

To try to get a better idea, I tracerout6ed to a 6to4 address using
a routing header to get the packet to first go to nodes that had
had replies encapsulated with the anycast address. The second last
hop of these traceroutes should help identify the 6to4 relay.

This produced 27 IPv6 addresses, but sevral of them looked like
they were assigned to the same machine. Looking at the /32s that
these IPv6 addresses are in suggests that there are actually about
12 distinct relays in this group. Looking them up in whois shows
that two of the /32s are in switch, so we're probably looking at
11 (CH:1 DE:3 EE:1 FI:1 IE:1 JP:1 NL:1 UK:1 US:1).

In total, that's about 33, if you take into account a possible
overlap of 2 or 3.

[Naturally, I may have missed lots of relays that aren't visable
from the route to the ::1 address within a prefix...]

        David.