And now, for your regularly scheduled dose of broken ipv6 routing...

[Carlos Friacas]

On Sun, 24 Sep 2006, Iljitsch van Beijnum wrote:

> On 22-sep-2006, at 10:16, Carlos Friacas wrote:
>
>> Recently i've started to measure v4 and v6 latency to the same host (or to 
>> the same name with both A and AAAA records) and found that it's really, 
>> really hard to find names with both A and AAAA in some countries (and yes, 
>> *even* in the academic communities...)
>
>> For what i've seen so far, the measurements i get with a smaller difference 
>> than 1ms between v4 & v6 are from other NRENs (to be more accurate... 13 of 
>> them to the date). Could be better, could be worse...
>
> I have to disagree here.

not sure where exactly we disagree...


> It all depends on your vantage point. Within the 
> commercial IPv6 internet, the difference between IPv4 and IPv6 is probably a 
> bit bigger than within the academic networks, BUT... apparently, the academic 
> networks as a group refuse to take good peering

afaik, yes.
the current policy is to peer with other academic networks, and to buy 
transit to reach the rest of the world.
i have to live with it, despite thinking its a good or bad one...

the only thing i can do atm is to peer with other existing v6 networks 
accross the local exchange.


> and transit to the commercial 
> world in IPv6, so going between the two sucks big time in many cases.

strongly agree :-(((
i especially dislike to see europe->europe v6 flows going through 
trans-continental links...


> I was going to show you that from my system to www.fccn.pt IPv6 is worse than 
> IPv4, but I don't get any ping replies over IPv4 so never mind...

:-) same problem i felt when trying to find v4+v6 destinations.
this is something yet to be tuned... v6 policy should match the v4 
one, so one day ping6 will no longer work too ;-)

but anyone can use our 193.136.5.1 and 2001:690::1 addresses for testing 
purposes related with AS1930 - it's the same box.


> I'll be interested to see how this one is for you:

sure, see below.


> [alumange:~] iljitsch% ping -c 25 www.isc.org
> PING www.isc.org (204.152.184.88): 56 data bytes
> 25 packets transmitted, 25 packets received, 0% packet loss
> round-trip min/avg/max/stddev = 168.561/187.332/250.520/17.626 ms

PING www.isc.org (204.152.184.88) 56(84) bytes of data.
--- www.isc.org ping statistics ---
25 packets transmitted, 25 received, 0% packet loss, time 23999ms
rtt min/avg/max/mdev = 204.546/210.303/215.547/3.974 ms


> [alumange:~] iljitsch% ping6 -c 25 www.isc.org
> PING6(56=40+8+8 bytes) 2001:1af8:6::20a:95ff:fef5:246e --> 2001:4f8:0:2::d
> 25 packets transmitted, 25 packets received, 0% packet loss
> round-trip min/avg/max = 186.890/255.224/635.499 ms

PING www.isc.org(www.isc.org) 56 data bytes
--- www.isc.org ping statistics ---
25 packets transmitted, 25 received, 0% packet loss, time 24015ms
rtt min/avg/max/mdev = 223.542/223.796/224.226/0.606 ms, pipe 2


My "bad v6 routing factor"(TM) seems to be smaller than yours. :-))


Cheers,

./Carlos                                               Skype: cf916183694
--------------
          Wide Area Network (WAN) Workgroup, CMF8-RIPE, CF596-ARIN
FCCN - Fundacao para a Computacao Cientifica Nacional  http://www.fccn.pt

  "Internet is just routes (196663/675), naming (millions) and... people!"