Wanted: Abilene Connectivity

Bernhard Schmidt berni at birkenwald.de
Wed Jan 4 02:28:15 CET 2006


Hi everyone,

I'm seeing heavy problems connecting to anything within Abilene's IPv6
prefix from well-connected networks lately (well, for several months now)

a) 29259 8767 3549 6939 293 11537

traceroute to hammer.uoregon.edu (2001:468:d01:dc::80df:dc1d) from
2001:1b10:100::1:1, 30 hops max, 16 byte packets
 1  backbone2-gige-0-1-16.teleport-iabg.de (2001:1b10:100:4::12)  0.341 ms
 2  fe-0-2-2.rt7.muc3.m-online.net (2001:a60:0:201::1:1)  9.593 ms
 3  so-1-2-0.rt9.muc1.m-online.net (2001:a60:0:100::1)  1.991 ms
 4  ge-1-1-0.408.ar1.FRA2.gblx.net (2001:a60:0:1ff::2)  13.809 ms
 5  2001:450:2001:1000:0:670:1708:219
(2001:450:2001:1000:0:670:1708:219)  172.887 ms
 6  *
 7  *
 8  *
 9  *

Dies at GBLX->HE, various guys in IRC have confirmed the same. Since
other 3549 6939 !293 paths work it might be a problem between HE and
ES.net. But

b) 29259 286 1273 6830 6830 6830 6830 6939 293 11537

 1  backbone2-gige-0-1-16.teleport-iabg.de (2001:1b10:100:4::12)  0.673 ms
 2  mchn-s2-rou-1030.DE.eurorings.net (2001:680:0:8008::2)  2.546 ms
 3  ge-0-1-0-gar1.muc.cw.net (2001:5001:100:15::1)  2.518 ms
 4  ge-0-0-0-200-zcr2.muc.cw.net (2001:5000:0:4::2)  3.461 ms
 5  as0-bcr2.fra.cw.net (2001:5000:0:5::1)  13.002 ms
 6  so-1-2-0-dcr2.fra.cw.net (2001:5000:0:f::1)  14.8 ms
 7  so-4-0-0-dcr1.amd.cw.net (2001:5000:0:e::2)  21.82 ms
 8  so-4-0-0-bcr1.amd.cw.net (2001:5000:0:10::2)  22.159 ms
 9  so-3-0-0-zcr1.amt.cw.net (2001:5000:0:12::2)  22.477 ms
10  nl-ams04a-re1-fe-0-0.ipv6.aorta.net (2001:7f8:1::a500:6830:1)  26.293 ms
11  nl-ams06d-re1-t-2.ipv6.aorta.net (2001:730::1:c)  29.711 ms
12  hurrican.net-gw1.nl.ipv6.aorta.net (2001:730::1:2f)  181.936 ms
13  3ffe:81d0:ffff:1::1 (3ffe:81d0:ffff:1::1)  182.213 ms
14  3ffe:80a::a (3ffe:80a::a)  290.199 ms
15  snvrt1-paixpav6r1.es.net (2001:400:0:50::1)  290.679 ms
16  2001:468:ff:17c3::1 (2001:468:ff:17c3::1)  289.534 ms
17  oregon-snvang.abilene.ucaid.edu (2001:468:ff:174d::2)  303.009 ms
18  2001:468:d00:a390::3 (2001:468:d00:a390::3)  303.619 ms
19  ge-5-1.uonet1-gw.uoregon.edu (2001:468:d01:2::1)  302.761 ms
20  hammer.ipv6.uoregon.edu (2001:468:d01:dc::80df:dc1d)  302.923 ms

works, although it shares 6939 293. Different peering between HE and
ES.net due to different ingress interfaces?

c) no as-path, but a trace from my colo-box:

 2  2001:1b18:f:16::1 (2001:1b18:f:16::1)  7.652 ms
 3  decix1.ipv6.eurotransit.net (2001:7f8::73f6:0:1)  7.576 ms
 4  so-2-0-0.pr1.k90.fra.de.v6.eurotransit.net (2001:1bc0::ffff:ffff:11)
 8.015 ms
 5  Ge7-0-1.FFTCR2.Frankfurt.opentransit.net (2001:688:0:3:7::9)  26.334 ms
 6  Po11-0.FFTCR3.Frankfurt.opentransit.net (2001:688:0:2:1::1)  8.307 ms
 7  Po10-0.AUVCR2.Aubervilliers.opentransit.net (2001:688:0:2:1::19)
21.398 ms
 8  Po6-0.NYKCR2.New-york.opentransit.net (2001:688:0:2:8::2)  98.568 ms
 9  Po13-0.CHICR2.Chicago.opentransit.net (2001:688:0:2:1::e)  117.566 ms
10  Abilene-1-ipv6.GW.opentransit.net (2001:688:0:4::19)  280.154 ms
11  iplsng-chinng.abilene.ucaid.edu (2001:468:ff:f12::2)  280.058 ms
12  kscyng-iplsng.abilene.ucaid.edu (2001:468:ff:1213::2)  291.715 ms
13  dnvrng-kscyng.abilene.ucaid.edu (2001:468:ff:1013::1)  287.211 ms
14  snvang-dnvrng.abilene.ucaid.edu (2001:468:ff:1017::2)  280.023 ms
15  oregon-snvang.abilene.ucaid.edu (2001:468:ff:174d::2)  296.637 ms
16  2001:468:d00:a390::3 (2001:468:d00:a390::3)  293.08 ms
17  ge-5-1.uonet1-gw.uoregon.edu (2001:468:d01:2::1)  294.428 ms
18  hammer.ipv6.uoregon.edu (2001:468:d01:dc::80df:dc1d)  292.173 ms

okay, latency sucks (so probably the path back goes through Asia), but
at least we get there through a more or less sane path.

I've seen reports of _well-connected networks_ having no less than 600ms
to various Abilene hosts (hammer.uoregon.edu tested, but they all share
the same pain). It used to be better, I remember using both C&W and GBLX
direct peerings with Abilene, can anyone say what happened to them? The
overview of the paths to the abilene prefix in grh.sixxs.net looks
hillariously broken, with ASes filtered for having a crappy routing
policy (like HE.net) or Asian ASes like 2500 and 4725 all over the
place. A small collection

29670 12732 20646 1752 1299 3320 5511 11537 i
24875 9009 3257 2497 2500 4725 11537 i
12779 6175 6830 6830 6830 6939 293 11537 i

The only point where I do get good latency is the german NREN connected
to GEANT, who have a direct "something" with Abilene. Anyone else having
the same issues? Is this intended or is something broken here?

Regards,
Bernhard


More information about the ipv6-ops mailing list