High latency to tunnel server in Ashburn from cogent-based Ashburn connection?

[pc500]

This used to be <10ms last time I used it a while back.  Now it's deciding to go to San Jose and back.  

   te2-2.mpd01.iad01.atlas.cogentco.com (154.54.26.94) 84 msec 0 msec
 3 te0-1-0-0.mpd21.iad02.atlas.cogentco.com (154.54.31.113) 0 msec
   te0-0-0-6.mpd21.iad02.atlas.cogentco.com (154.54.31.205) 0 msec
   te0-1-0-6.mpd21.iad02.atlas.cogentco.com (154.54.31.161) 0 msec
 4 te2-3-10GE.ar5.DCA3.gblx.net (64.212.107.89) 0 msec 0 msec 4 msec
 5 po3-20G.ar3.SJC2.gblx.net (67.16.141.62) 68 msec 72 msec 72 msec
 6 Hurrican-Electric-LLC.Port-channel100.ar3.SJC2.gblx.net (64.214.174.246) 68 msec 76 msec 72 msec
 7 10gigabitethernet3-2.core1.pao1.he.net (72.52.92.69) 72 msec 76 msec 72 msec
 8 10gigabitethernet2-4.core1.ash1.he.net (72.52.92.30) 72 msec 72 msec
   10gigabitethernet9-3.core1.ash1.he.net (184.105.213.178) 72 msec
 9 ash.ipv6.he.net (216.218.229.118) 72 msec *  68 msec

While I can't check a return traceroute directly from ash.ipv6.he.net -- I can from your looking glass ashburn entry.  The return has proper routing (I am basing this by the latency in the reply, not the output of the traceroute as is is assumed to be asymmetrical for now).  But perhaps it's sourcing the traceroute from an interface address on telia?  Maybe this explains the quick replies as I can pine that hop#1 very quickly too?  I don't know -- I have no easy way to find the source IP of the looking glass traceroute, and the LG provides only dns hostnames.

core1.ash1.he.net> traceroute x.x.x.x numeric

Hop    Packet 1    Packet 2    Packet 3    Hostname
1    <1 ms    <1 ms    <1 ms    ash-bb1-link.telia.net (213.248.67.117)
2    <1 ms    <1 ms    <1 ms    te0-7-0-2.mpd21.iad02.atlas.cogentco.com (154.54.10.57)
3    <1 ms    <1 ms    <1 ms    te0-0-0-0.mpd22.iad02.atlas.cogentco.com (154.54.31.110)
4    <1 ms    <1 ms    <1 ms    te2-7.ccr02.iad01.atlas.cogentco.com (154.54.31.214)
5    <1 ms    <1 ms    11 ms    x.x.x.x

I even try tracerouting to core1.ash1.he.net and can not produce the same results.  Or maybe the DNS entry doesn't match the looking glass source IP?

Tracing the route to core1.ash1.he.net (216.218.252.169)

 1 x.cogentco.com (38.104.) 0 msec 0 msec 0 msec
 2 te0-1-0-6.mpd22.iad02.atlas.cogentco.com (154.54.31.213) 0 msec
   te0-0-0-6.ccr22.iad02.atlas.cogentco.com (154.54.31.193) 0 msec
   te2-2.mpd01.iad01.atlas.cogentco.com (154.54.26.94) 0 msec
 3 te0-2-0-0.mpd21.iad02.atlas.cogentco.com (154.54.31.126) 0 msec 0 msec
   te0-1-0-6.mpd21.iad02.atlas.cogentco.com (154.54.31.161) 4 msec
 4 te2-3-10GE.ar5.DCA3.gblx.net (64.212.107.89) 0 msec 0 msec 0 msec
 5 po3-20G.ar3.SJC2.gblx.net (67.16.141.62) 72 msec 68 msec 72 msec
 6 Hurrican-Electric-LLC.Port-channel100.ar3.SJC2.gblx.net (64.214.174.246) 72 msec 68 msec 72 msec
 7 10gigabitethernet3-2.core1.pao1.he.net (72.52.92.69) 72 msec 72 msec 72 msec
 8  *  *  *
 9  *

[broquea]

You should ask your provider Cogent to either speak to GBLX or use Telia instead since that seems fine. No reason even getting to the core should be going to SJC first. Maybe GBLX can hand it to Cogent in Ashburn again. Maybe this is how Cogent wants it. Again, check with your provider.

[pc500]

Thank you, I will ask them.  If there is a connection from GBLX to HE locally as you indicate, that's what it should take then.

[pc500]

There response was:
"Per our NOC investigation, we cannot change the routing.  The route originates in VA and we hand it off in VA.  Global Crossing learns the route in CA.  Routing will be dependent on how and where the route is being advertised."

Basically what they seem to be saying is GBLX only has an interconnect with HE in California... so it has no choice but to go there.

[houkouonchi]

I got bitten by this as well. Luckily it doesn't affect me as badly because I am in Southern California (but routes through northern california, when it used to be direct). It is also global crossing =(

root@dekabutsu: 06:28 AM :~# traceroute -I  66.220.18.41
traceroute to gige-g4-6.core1.lax1.he.net (66.220.18.41), 30 hops max, 46 byte packets
1  router.houkouonchi.jp (1.1.1.1)  0.284 ms  0.408 ms  0.168 ms
2  L103.LSANCA-DSL-29.verizon-gni.net (71.110.63.1)  1.033 ms  1.060 ms  0.807 ms
3  G0-3-1-7.LSANCA-LCR-22.verizon-gni.net (130.81.146.60)  1.437 ms  1.448 ms  1.286 ms
4  so-6-0-0-0.LAX01-BB-RTR2.verizon-gni.net (130.81.29.126)  2.039 ms  1.995 ms  2.504 ms
5  0.ae2.BR3.LAX15.ALTER.NET (152.63.2.133)  2.272 ms  2.278 ms  2.250 ms
6  te6-3-10G.ar4.LAX2.gblx.net (64.208.110.169)  2.472 ms  2.396 ms  2.274 ms
7  Hurrican-Electric-LLC.Port-channel100.ar3.SJC2.gblx.net (64.214.174.246)  24.172 ms  24.079 ms  25.058 ms
8  * * *
9  * * *
10  tserv15.lax1.ipv6.he.net (66.220.18.42)  13.863 ms  13.830 ms  13.846 ms

I used to get like <5ms latency.. It was nice.