Thanks for the post Mirjam Regarding latency spike - 300ms is a nice number. This is average latency between India and US and I am quite sure STPI AS9430 is not hearing BSNL AS9829 blocks in India (just like many other ISP's) due to misconfiguration from BSNL at that time. STPI is using Tata Comm, Reliance & Airtel for upstream. Most likely they somehow were not exchanging traffic in niXi with BSNL. In general BSNL seems to be having very much isolated in terms of peering. They peer with multiple tier 1 - Tata, Gblx, At&t but it's seems like more of transit capacity purchase then real peering. Forward to paths to Europe and Singapore looks pretty good but reverse path is very very poor due to fact that most of packets return after hitting New York or Los Angles. E.g forward path to openDNS who is doing anycasting and nearest node is in Singapore: anurag@laptop:~$ traceroute 208.67.222.222 -A traceroute to 208.67.222.222 (208.67.222.222), 30 hops max, 60 byte packets 1 router.local (192.168.1.1) [AS8151/AS28513] 2.518 ms 3.409 ms 4.115 ms 2 117.206.176.1 (117.206.176.1) [AS9829] 27.053 ms 28.224 ms 29.389 ms 3 218.248.169.122 (218.248.169.122) [AS9829] 35.495 ms 38.114 ms 39.288 ms 4 115.114.57.165.static-Mumbai.vsnl.net.in<http://115.114.57.165.static-mumbai.vsnl.net.in/> (115.114.57.165) [AS4755] 76.229 ms 77.389 ms 78.378 ms 5 172.31.16.174 (172.31.16.174) [*] 102.700 ms 103.997 ms 106.173 ms 6 ix-4-2.tcore2.CXR-Chennai.as6453.net<http://ix-4-2.tcore2.cxr-chennai.as6453.net/> (180.87.37.1) [*] 223.402 ms 203.658 ms 206.171 ms 7 if-5-2.tcore2.SVW-Singapore.as6453.net<http://if-5-2.tcore2.svw-singapore.as6453.net/> (180.87.15.69) [*] 239.135 ms 240.445 ms 242.632 ms 8 Vlan1807.icore1.SVQ-Singapore.as6453.net<http://vlan1807.icore1.svq-singapore.as6453.net/> (120.29.215.29) [*] 245.485 ms 259.768 ms 260.764 ms 9 203.208.186.101 (203.208.186.101) [AS7473] 407.914 ms 409.414 ms 410.124 ms 10 203.208.153.110 (203.208.153.110) [AS7473] 405.858 ms 406.540 ms 408.564 ms 11 203.208.190.166 (203.208.190.166) [AS7473] 412.533 ms 413.239 ms 413.928 ms 12 * * * 13 * * * 14 * * * 400ms latency, while it should be as low as 120-130ms. Forward path from my system to openDNS is like: BSNL AS9829 >> AS4755-Tata >> AS6453-Tata >> AS7473-Singtel >> OpenDNS-AS36692 Let's skip AS9829 and checking same route from next network - AS4755 from their looking glass: Router: TATA COMM DELHI(AS4755) Site: TATA COMM DELHI(AS4755) Command: traceroute 208.67.222.222 as-number-lookup traceroute to 208.67.222.222 (208.67.222.222) from 121.244.139.224, 30 hops max, 40 byte packets 1 172.31.17.5 (172.31.17.5) 56.232 ms 49.907 ms 49.865 ms 2 ix-4-2.tcore1.CXR-Chennai.as6453.net<http://ix-4-2.tcore1.cxr-chennai.as6453.net/> (180.87.36.9) 50.788 ms 49.942 ms 49.950 ms 3 if-3-3.tcore2.CXR-Chennai.as6453.net<http://if-3-3.tcore2.cxr-chennai.as6453.net/> (180.87.36.6) [AS 6453] 82.654 ms if-5-2.tcore1.SVW-Singapore.as6453.net<http://if-5-2.tcore1.svw-singapore.as6453.net/> (180.87.12.53) [AS 6453] 83.601 ms if-3-3.tcore2.CXR-Chennai.as6453.net<http://if-3-3.tcore2.cxr-chennai.as6453.net/> (180.87.36.6) [AS 6453] 82.719 ms MPLS Label=464996 CoS=0 TTL=1 S=1 4 if-2-2.tcore2.SVW-Singapore.as6453.net<http://if-2-2.tcore2.svw-singapore.as6453.net/> (180.87.12.2) [AS 6453] 106.121 ms 90.087 ms 82.476 ms 5 Vlan1807.icore1.SVQ-Singapore.as6453.net<http://vlan1807.icore1.svq-singapore.as6453.net/> (120.29.215.29) [AS 6453] 89.501 ms 87.419 ms 89.912 ms 6 203.208.186.101 (203.208.186.101) [AS 7473] 85.556 ms 83.610 ms 91.788 ms 7 ge-3-0-0-0.sngtp-dr2.ix.singtel.com (203.208.151.157) [AS 7473] 84.276 ms 103.993 ms 84.523 ms 8 203.208.190.166 (203.208.190.166) [AS 7473] 83.971 ms 84.105 ms 85.290 ms 9 * * * 10 * * * Good latency - thus clearly issue is NOT with Tata Comm here. Looking at reverse path from 2nd network on route i.e Singtel to BSNL from their looking glass: traceroute 117.206.176.37 Type escape sequence to abort. Tracing the route to 117.206.176.37 1 203.208.155.2 0 msec 4 msec 0 msec 2 ge-0-1-2-0.sngc3-dr1.ix.singtel.com (203.208.172.117) 0 msec ge-1-1-4-0.sngc3-dr1.ix.singtel.com (203.208.172.169) 0 msec 0 msec 3 203.208.153.245 4 msec ge-1-1-3-0.sngtp-dr2.ix.singtel.com (203.208.152.21) 4 msec 203.208.152.229 0 msec 4 so-2-0-2-0.laxow-cr1.ix.singtel.com (203.208.151.118) 180 msec 203.208.178.186 221 msec 203.208.154.46 184 msec 5 so-2-0-3-0.plapx-cr2.ix.singtel.com (203.208.149.246) 184 msec 189 msec 184 msec 6 ge-2-0-0-0.sngc3-dr1.ix.singtel.com (203.208.183.38) 180 msec 180 msec 181 msec 7 if-2-2.tcore2.LVW-LosAngeles.as6453.net (66.110.59.2) [MPLS: Label 433504 Exp 0] 336 msec 345 msec 337 msec 8 if-1-3.tcore2.NJY-Newark.as6453.net (64.86.252.58) [MPLS: Label 337184 Exp 0] 328 msec 349 msec 328 msec 9 if-2-2.tcore1.NJY-Newark.as6453.net (66.198.70.1) [MPLS: Label 341104 Exp 0] 349 msec 324 msec 321 msec 10 if-4-2.tcore1.L78-London.as6453.net (80.231.130.33) [MPLS: Label 619459 Exp 0] 332 msec 329 msec 337 msec 11 if-13-0-0-2.mcore3.L78-London.as6453.net (80.231.130.30) 336 msec if-0-0-0.mcore3.L78-London.as6453.net (80.231.130.50) 329 msec if-13-0-0-2.mcore3.L78-London.as6453.net (80.231.130.30) 336 msec 12 ix-2-0-1.mcore3.L78-London.as6453.net (195.219.144.78) 253 msec 256 msec 252 msec 13 218.248.255.89 245 msec 256 msec 256 msec 14 218.248.169.121 309 msec 293 msec 296 msec 15 218.248.169.121 293 msec 297 msec 296 msec 16 * * * Similar problems exist for Europe too. (not putting data to keep email short, let me know if you need it). Thus at the end "ring" is between India & US for packets even when server is in Singapore. Missing k-root node not only increases latency but also puts a complete 360 degree round trip to US. Thanks. On Tue, Feb 21, 2012 at 2:31 PM, Mirjam Kuehne <mir@ripe.net> wrote:
Hi,
As a follow-up, we looked at the connectivity at the K-root instance in Delhi. Please find a short analysis on RIPE Labs:
https://labs.ripe.net/Members/**emileaben/connectivity-to-the-** k-root-instance-in-Delhi<https://labs.ripe.net/Members/emileaben/connectivity-to-the-k-root-instance-in-Delhi>
Kind Regards, Mirjam Kuehne RIPE NCC
On 17/2/12 5:12 PM, Anurag Bhatia wrote:
Nice report Mirjam
Thanks for sharing. I found k-root to be having issues in India multiple times. Sometimes due to downtime of Delhi node, other time due to messed up routing which takes off traffic outside India, increasing latency from 30ms to over 150ms.
I am curious to see RIPE probe collected data for the k-root.
On Fri, Feb 17, 2012 at 9:02 PM, Mirjam Kuehne <mir@ripe.net <mailto:mir@ripe.net>> wrote:
[apologies for duplicate mails]
Dear colleagues,
Please see this new article on RIPE Labs describing a tool that visualises the performance of K-root:
https://labs.ripe.net/Members/**__csquarce/visualk<https://labs.ripe.net/Members/__csquarce/visualk>
<https://labs.ripe.net/**Members/csquarce/visualk<https://labs.ripe.net/Members/csquarce/visualk>
Kind regards, Mirjam Kuehne RIPE NCC
--
Anurag Bhatia anuragbhatia.com <http://anuragbhatia.com>
or simply - http://[2001:470:26:78f::5] if you are on IPv6 connected network!
Twitter: @anurag_bhatia <https://twitter.com/#!/**anurag_bhatia<https://twitter.com/#!/anurag_bhatia>
Linkedin: http://linkedin.anuragbhatia.**com<http://linkedin.anuragbhatia.com>
-- Anurag Bhatia anuragbhatia.com or simply - http://[2001:470:26:78f::5] if you are on IPv6 connected network! Twitter: @anurag_bhatia <https://twitter.com/#!/anurag_bhatia> Linkedin: http://linkedin.anuragbhatia.com