poor routes from Road Runner on east coast

[kjotte]

I am on Earthlink (thus effectively RR) in North Carolina, so one would think that ASH or NYC (maybe even MIA) would be good choices.  Unfortunately no matter what tunnel server I trace to, I get hauled across RR's backbone to a peering point in San Jose.  Someone here in the forums was having a similar problem out of DAL about a year ago, but I don't think anything ever came of it.  Knowing full well that they won't listen to a lowly customer like me about things like peering, is there some other avenue to improve my route?

Some traces for reference:

Code Select Expand


kjotte@polaris:~$ tracepath tserv7.ash1.ipv6.he.net
1:  polaris-wifi.nivex.lan (172.31.3.23)                   0.221ms pmtu 1500
1:  peregrine.nivex.lan (172.31.3.254)                     2.143ms
1:  peregrine.nivex.lan (172.31.3.254)                     1.855ms
2:  user-0c2h381.cable.mindspring.com (24.40.141.1)       16.944ms
3:  gig0-1-1.rlghnca-rtr2.nc.rr.com (66.26.33.157)        34.706ms
4:  xe-4-0-1.chrlncpop-rtr1.southeast.rr.com (24.93.64.6)  45.202ms asymm  5
5:  ae-3-0.cr0.atl20.tbone.rr.com (66.109.6.82)           41.377ms asymm  6
6:  ae-3-0.cr0.dfw10.tbone.rr.com (66.109.6.36)           45.044ms
7:  ae-0-0.cr0.hou30.tbone.rr.com (66.109.6.38)           56.993ms
8:  ae-3-0.cr0.lax30.tbone.rr.com (66.109.6.0)            90.059ms
9:  ae-0-0.cr0.lax00.tbone.rr.com (66.109.6.4)            83.591ms
10:  ae-3-0.cr0.sjc30.tbone.rr.com (66.109.6.6)            91.761ms
11:  ae-1-0.pr0.sjc10.tbone.rr.com (66.109.6.137)         135.536ms asymm 12
12:  gige-g5-6.core1.sjc2.he.net (216.218.135.225)         91.988ms asymm 13
13:  10gigabitethernet3-2.core1.pao1.he.net (72.52.92.69)  92.059ms asymm 12
14:  10gigabitethernet2-4.core1.ash1.he.net (72.52.92.30) 104.432ms asymm 10
15:  ash.ipv6.he.net (216.218.229.118)                    106.899ms reached
    Resume: pmtu 1500 hops 15 back 245
kjotte@polaris:~$ tracepath tserv4.nyc4.ipv6.he.net
1:  polaris-wifi.nivex.lan (172.31.3.23)                   0.176ms pmtu 1500
1:  peregrine.nivex.lan (172.31.3.254)                     1.727ms
1:  peregrine.nivex.lan (172.31.3.254)                     1.806ms
2:  user-0c2h381.cable.mindspring.com (24.40.141.1)       16.393ms
3:  gig0-1-1.rlghnca-rtr2.nc.rr.com (66.26.33.157)        38.942ms
4:  ge-5-3-0.chrlncpop-rtr1.southeast.rr.com (24.93.64.2)  43.296ms asymm  5
5:  ae-3-0.cr0.atl20.tbone.rr.com (66.109.6.82)           72.009ms asymm  6
6:  ae-3-0.cr0.dfw10.tbone.rr.com (66.109.6.36)           56.382ms
7:  ae-0-0.cr0.hou30.tbone.rr.com (66.109.6.38)           52.791ms
8:  ae-3-0.cr0.lax30.tbone.rr.com (66.109.6.0)            81.374ms
9:  ae-0-0.cr0.lax00.tbone.rr.com (66.109.6.4)            84.013ms
10:  ae-3-0.cr0.sjc30.tbone.rr.com (66.109.6.6)           107.589ms
11:  ae-1-0.pr0.sjc10.tbone.rr.com (66.109.6.137)          91.693ms asymm 12
12:  gige-g5-15.core1.sjc2.he.net (216.218.135.229)       105.623ms asymm 13
13:  10gigabitethernet1-3.core1.nyc4.he.net (72.52.92.26) 108.806ms
14:  no reply
15:  no reply
^C


[comptech]

I've got the same problem in Kansas City.  Road Runner appears to be routing everything destined for HE to San Jose for some reason.
I don't know what we can do to get it fixed since it looks like it's Road Runner's routing that has a problem.  I imagine all Road Runner customers are getting terrible routes to their tunnel servers because of this, unless you're lucky enough to be near San Jose.

Code Select Expand

4:  RDC-24-94-161-34.kc.rr.com (24.94.161.34)             29.557ms asymm  5
5:  ae-5-0.cr0.chi30.tbone.rr.com (66.109.6.112)          32.310ms asymm  6
6:  ae-0-0.cr0.chi10.tbone.rr.com (66.109.6.20)           29.395ms asymm  7  <- Passing through Chicago!
7:  ae-3-0.cr0.sjc10.tbone.rr.com (66.109.6.14)           80.890ms asymm  8
8:  ae-0-0.pr0.sjc10.tbone.rr.com (66.109.6.141)          82.254ms asymm 10
9:  gige-g5-15.core1.sjc2.he.net (216.218.135.229)        95.587ms asymm 14
10:  10gigabitethernet1-1.core1.chi1.he.net (72.52.92.74)  84.965ms asymm  9
11:  tserv9.chi1.ipv6.he.net (209.51.181.2)                85.874ms reached


Traceroute to one hop before the Chicago HE core:

Code Select Expand

 
                                   Loss%   Snt   Last   Avg  Best  Wrst StDev
  4. CPE-69-76-62-253.kc.res.rr.c  0.0%    10   19.8  21.4  19.8  26.1   2.0
  5. ae-4-0.cr0.chi10.tbone.rr.co  0.0%    10   20.5  21.6  19.9  29.3   2.9
  6. ae-0-0.pr0.chi10.tbone.rr.co  0.0%    10   25.0  29.5  24.1  55.1   9.5
  7. chi-bb1-link.telia.net        0.0%    10   20.3  26.4  18.7  57.1  11.5

Traceroute to the Chicago HE core:

Code Select Expand

 
                                   Loss%   Snt   Last   Avg  Best  Wrst StDev
  4. CPE-69-76-62-253.kc.res.rr.c  0.0%    10   39.5  24.1  19.5  39.5   6.0
  5. ae-4-0.cr0.chi10.tbone.rr.co  0.0%    10   20.4  21.5  19.3  26.5   2.1
  6. ae-3-0.cr0.sjc10.tbone.rr.co  0.0%    10   75.3  80.6  75.3  92.4   6.3
  7. ae-0-0.pr0.sjc10.tbone.rr.co  0.0%    10   79.7  83.1  77.1 117.1  12.0
  8. gige-g5-6.core1.sjc2.he.net   0.0%    10   86.3  87.1  84.8  90.0   1.7
  9. core1.chi1.he.net             0.0%    10   77.2  84.4  77.2 100.0   7.6

Traceroute from Chicago HE core to me:

Code Select Expand

1  5 ms  <1 ms  <1 ms  chi-bb1-link.telia.net (213.248.104.213)
2 <1 ms <1 ms <1 ms timewarner-ic-130992-chi-bb1.c.telia.net (213.248.76.98)
3 <1 ms <1 ms 1 ms ae-1-0.cr0.chi30.tbone.rr.com (66.109.6.154)
4 <1 ms <1 ms <1 ms ae-15.chcgill3-rtr1.kc.rr.com (66.109.6.113)
5 20 ms <1 ms 23 ms RDC-24-94-161-33.kc.rr.com (24.94.161.33)

[kjotte]

Oh joy, the routing is asymmetric.  I just traced from ASH back to me:

Code Select Expand


core1.ash1.he.net> traceroute 24.40.141.158 numeric
1  <1 ms  <1 ms  <1 ms  ash-bb1-link.telia.net (213.248.98.93)
2 <1 ms <1 ms <1 ms timewarner-ic-130990-ash-bb1.c.telia.net (213.248.89.194)
3 <1 ms <1 ms <1 ms ae-2-0.cr0.dca10.tbone.rr.com (66.109.6.168)
4 16 ms 16 ms 16 ms ae-1-0.cr1.atl20.tbone.rr.com (66.109.6.32)
5 17 ms 16 ms 17 ms ae-0-0.cr0.atl20.tbone.rr.com (66.109.6.34)
6 21 ms 21 ms 21 ms 66.109.6.83
7 26 ms 26 ms 26 ms gig4-0-0.rlghnca-rtr2.southeast.rr.com (24.93.64.177)
8 28 ms 27 ms 28 ms gig16-1.rlghnce-ar42.nc.rr.com (66.26.33.158)
9 45 ms 48 ms 50 ms user-0c2h3cu.cable.mindspring.com (24.40.141.158)

That's a route I could live with.

[brad]

Oh joy, the routing is asymmetric.

This is the Internet. That has been the norm for a long time.

[snarked]

As the routing is assymetric, the "problem" lies with RoadRunner and their importing of routes external to their network into their own routers at the BGP level.  Apparently, they're only importing a route to HE from their peering with HE in California and not accepting routes to HE via other carriers (even when those routes are shorter).

[kjotte]

So I guess the question is: Do I go ahead and redo my tunnel for Ashburn, despite the recommendation of the tunnel creation page, knowing that my return packets will be taking a much shorter path?

[brad]

So I guess the question is: Do I go ahead and redo my tunnel for Ashburn, despite the recommendation of the tunnel creation page, knowing that my return packets will be taking a much shorter path?

You can have tunnels wherever you like. The recommendation is just that, a recommendation. If having the tunnel to the Ashburn POP results in better latency with your connection then by all means setup a tunnel to that POP.