Europe-to-US congestion and packet loss on he.net network, and their NOC@ won't even respond
Constantine A. Murenin
mureninc at gmail.com
Sun Dec 1 09:11:54 UTC 2013
On 2013-W48-6 23:19 -0800, Matthew Petach wrote:
> On Fri, Nov 29, 2013 at 11:30 PM, Constantine A. Murenin <mureninc at gmail.com
> > wrote:
>
> > Dear NANOG@,
> >
>
> ...
>
>
> > From hetzner.de through he.net:
> >
> >
> > Cns# date ; mtr --report{,-wide,-cycles=600} --interval 0.1 --order "SRL
> > BGAWV" -4 ????c????????.indiana.edu ; date
> >
>
>
> Using a 1/10th of a second interval is rather anti-social.
> I know we rate-limit ICMP traffic down, and such a
> short interval would be detected as attack traffic,
> and treated as such.
>
> I would take any results you get from such probes
> with a grain of salt. What results do you get with
> a more sane interval, one of at least 1 second or
> more?
>
> Matt
For what it is worth, I used to think the same, until I saw several
providers themselves suggest that 1000 packets should be sent, with
the 0.1 s interval. So, this is considered normal and appropriate
nowadays.
Anyhow, is this better?
I now saw a 2% traffic loss this night at a random test time, and
the 151ms avg rtt on this 114ms rtt route.
Cns# date ; mtr --report{,-wide,-cycles=600} --interval 0.5 --order "SRL BGAWV" -4 ????c????????.indiana.edu ; date
Sat Nov 30 23:17:13 PST 2013
HOST: Cns??????? Snt Rcv Loss% Best Gmean Avg Wrst StDev
1.|-- static.??.???.4.46.clients.your-server.de 600 600 0.0% 0.5 1.0 1.3 4.6 1.1
2.|-- hos-tr1.juniper1.rz13.hetzner.de 600 600 0.0% 0.1 0.2 2.0 58.5 7.9
3.|-- core21.hetzner.de 600 600 0.0% 0.2 0.2 0.2 10.2 0.7
4.|-- core22.hetzner.de 600 600 0.0% 0.2 0.2 0.2 11.2 0.8
5.|-- core1.hetzner.de 600 600 0.0% 4.8 4.8 4.8 25.1 1.3
6.|-- juniper1.ffm.hetzner.de 600 600 0.0% 4.8 4.8 4.8 13.9 0.6
7.|-- 30gigabitethernet1-3.core1.ams1.he.net 600 595 0.8% 11.2 14.3 15.2 121.4 7.4
8.|-- 10gigabitethernet1-4.core1.lon1.he.net 600 600 0.0% 18.2 21.0 21.3 51.2 4.0
9.|-- 10gigabitethernet10-4.core1.nyc4.he.net 600 592 1.3% 86.9 125.9 126.4 160.7 10.6
10.|-- 100gigabitethernet7-2.core1.chi1.he.net 600 591 1.5% 106.6 145.1 145.4 190.9 10.5
11.|-- ??? 600 0 100.0 0.0 0.0 0.0 0.0 0.0
12.|-- et-11-0-0.945.rtr.ictc.indiana.gigapop.net 600 589 1.8% 114.3 148.9 149.2 167.9 9.1
13.|-- xe-0-3-0.11.br2.ictc.net.uits.iu.edu 600 589 1.8% 113.4 149.2 149.5 173.4 9.3
14.|-- ae-0.0.br2.bldc.net.uits.iu.edu 600 590 1.7% 114.5 150.2 150.5 175.6 9.3
15.|-- ae-10.0.cr3.bldc.net.uits.iu.edu 600 589 1.8% 114.3 150.5 150.8 181.0 9.1
16.|-- ????c????????.indiana.edu 600 589 1.8% 114.8 150.7 151.0 170.7 9.0
Sat Nov 30 23:24:06 PST 2013
The ICMP timestamp request/reply test still indicates that only
one path is affected: the one from Europe to US over he.net.
Cns# date ; unbuffer hping --icmp-ts --count 30 ????c????????.indiana.edu | \
perl -ne 'if (/icmp_seq=(\d+) rtt=(\d+\.\d)/) {($s, $p) = ($1, $2);} \
if (/ate=(\d+) Receive=(\d+) Transmit=(\d+)/) {($o, $r, $t) = ($1, $2, $3);} \
if (/tsrtt=(\d+)/) { \
print $s, "\t", $p, "\t", $1, " = ", $r - $o, " + ", $o + $1 - $t, "\n"; }'
Sun Dec 1 00:55:46 PST 2013
0 151.3 151 = 91 + 60
1 154.2 154 = 93 + 61
2 127.8 127 = 67 + 60
3 123.6 123 = 63 + 60
4 136.9 137 = 76 + 61
5 149.6 149 = 89 + 60
6 147.4 147 = 87 + 60
7 133.5 133 = 73 + 60
8 152.2 152 = 92 + 60
9 137.3 137 = 77 + 60
10 143.7 144 = 84 + 60
11 124.5 124 = 64 + 60
12 141.4 141 = 81 + 60
13 118.0 118 = 58 + 60
14 153.6 154 = 94 + 60
15 137.7 138 = 78 + 60
16 119.9 120 = 60 + 60
17 130.6 131 = 71 + 60
18 144.6 145 = 85 + 60
19 138.8 139 = 79 + 60
20 155.7 156 = 96 + 60
21 128.8 129 = 69 + 60
22 153.0 153 = 93 + 60
23 146.5 147 = 87 + 60
24 137.2 138 = 77 + 61
25 153.3 154 = 94 + 60
26 146.3 147 = 87 + 60
27 150.1 151 = 91 + 60
28 150.5 150 = 90 + 60
29 143.5 143 = 83 + 60
Cheers,
Constantine.
More information about the NANOG
mailing list