[Update 2012-08-22 – 14:20]
IS fixed some local routing announces after my posting here. Nice to have a senior IS network engineer sitting next to me – Ta Riaan. As such latency to Telkom and MTN now looks better but capacity is still a tad ‘tight’.
So, sitting at the AFPIF3 meeting today and feeling like a train has parked on top of my head [Thanks winter flu – you can leave now!] and we are told about the amazing capacity (2Gbps available) to us at the conference.
As Neology is obviously active in the SA space I thought I’d do some quick basics tests and share them here:
First a traceroute from Afpif to Neology
traceroute to www.neology.co.za (41.73.40.129), 30 hops max, 60 byte packets 1 ge-1-1-9-0.br-01-jnb.mtg.afpif.org (196.216.207.253) 0.813 ms 2.612 ms 5.989 ms 2 41.87.104.217 (41.87.104.217) 5.929 ms 5.859 ms 5.776 ms 3 Gi9-0-0.jnb-201-access-3.mpls.seacomnet.com (41.217.213.130) 5.278 ms 5.200 ms 5.137 ms 4 mi-za-bry-gsrpe2-at0-0-51.ip.isnet.net (196.36.147.205) 5.078 ms 4.818 ms 4.682 ms 5 196.26.0.130 (196.26.0.130) 10.723 ms 10.709 ms 10.515 ms 6 * * * 7 196.43.25.137 (196.43.25.137) 192.466 ms 195.180 ms 195.185 ms 8 196.43.39.22 (196.43.39.22) 192.289 ms 195.187 ms 195.189 ms 9 196.25.223.253 (196.25.223.253) 3.103 ms 3.110 ms 3.111 ms 10 bnksr01-eth0.neoinx.net (41.216.193.11) 3.108 ms 1.425 ms 1.428 ms 11 hosting.neology.co.za (41.73.40.129) 3.229 ms 3.219 ms 3.231 ms
While there are some ‘funky’ hops along the line 3ms to our one hosting box makes reasonably good sense 😉
Lets try another location:
traceroute-nanog www.mtn.co.za traceroute to www.mtn.co.za (196.11.240.17), 30 hops max, 60 byte packets 1 ge-1-1-9-0.br-01-jnb.mtg.afpif.org (196.216.207.253) 13.379 ms 3.352 ms 4.313 ms 2 41.87.104.217 (41.87.104.217) 3.541 ms 5.494 ms 1.599 ms 3 Gi9-0-0.jnb-201-access-3.mpls.seacomnet.com (41.217.213.130) 3.208 ms 5.504 ms 5.145 ms 4 mi-za-bry-gsrpe2-at0-0-51.ip.isnet.net (196.36.147.205) 3.725 ms 4.575 ms 4.443 ms 5 196.26.0.130 (196.26.0.130) 16.283 ms 0.131 ms 0.201 ms 6 41-208-29-126.mtnns.net (41.208.29.126) 195.544 ms 194.259 ms 200.796 ms 7 jh-cr-2.za--jh-pr-1.za.mtnns.net (196.44.0.223) 198.400 ms 191.465 ms 200.890 ms 8 rb-cr-1.za--jh-cr-2.za-a.mtnns.net (196.44.31.94) 204.474 ms 205.415 ms 200.085 ms 9 rb-dca-1.za--rb-cr-1.za-a.mtnns.net (196.44.0.147) 198.919 ms 200.525 ms 197.219 ms 10 mtn-rb-1--rb-ad-1-a.mtnns.net (196.44.8.211) 203.558 ms 202.150 ms 200.767 ms 11 * * *
So while MTN firewalls ICMP we get a clear picture that MTN is ‘not’ considered local here — no peering. No shocks here – their peering is not exactly what we’d call ‘open’ ;-(
Now lets looks at MWEB:
traceroute-nanog www.mweb.co.za traceroute to www.mweb.co.za (196.2.63.110), 30 hops max, 60 byte packets 1 ge-1-1-9-0.br-01-jnb.mtg.afpif.org (196.216.207.253) 3.810 ms 2.845 ms 11.496 ms 2 41.87.104.217 (41.87.104.217) 3.693 ms 3.528 ms 3.402 ms 3 197-80-4-169.jhb.mweb.co.za (197.80.4.169) 3.603 ms 3.484 ms 3.281 ms 4 tengig0-0-0-1.vic-p-2.mweb.co.za (197.80.4.142) 27.324 ms 28.375 ms 27.285 ms 5 TenGigE0-0-0-0.cpt-p-1.mweb.co.za (197.84.4.34) 26.639 ms 27.883 ms 27.083 ms 6 vlan11.cpt-hscore-1.mweb.co.za (196.28.178.99) 26.342 ms 25.819 ms 26.007 ms 7 196.28.178.66 (196.28.178.66) 25.168 ms 16.271 ms 44.089 ms 8 cte-core-sw2.vwol.net (196.41.144.35) 18.838 ms 28.898 ms 32.171 ms 9 * * *
And Telkom:
traceroute-nanog www.telkom.co.za traceroute to www.telkom.co.za (196.43.22.222), 30 hops max, 60 byte packets 1 ge-1-1-9-0.br-01-jnb.mtg.afpif.org (196.216.207.253) 1.682 ms 1.657 ms 2.029 ms 2 41.87.104.217 (41.87.104.217) 12.427 ms 2.485 ms 0.815 ms 3 Gi9-0-0.jnb-201-access-3.mpls.seacomnet.com (41.217.213.130) 1.436 ms 0.657 ms 20.747 ms 4 mi-za-bry-gsrpe2-at0-0-51.ip.isnet.net (196.36.147.205) 3.376 ms 3.272 ms 4.725 ms 5 196.26.0.130 (196.26.0.130) 7.888 ms 3.647 ms 17.814 ms 6 * * * 7 196.43.25.137 (196.43.25.137) 195.097 ms 194.619 ms 195.302 ms 8 196.43.39.166 (196.43.39.166) 196.501 ms 194.647 ms 187.721 ms 9 rrba-ip-bssr-1-ge-2-48.telkom-ipnet.co.za (196.43.23.6) 201.266 ms 775.588 ms 708.520 ms 10 nbsc-ip-bssr-2-atm-5-0-0-1.telkom-ipnet.co.za (196.43.23.34) 726.548 ms 651.526 ms 207.583 ms 11 nbsc-ip-rcache-1-vif2.telkom-ipnet.co.za (196.43.22.222) 181.810 ms 188.047 ms 194.465 ms
IS fixed the announces to Telkom above – so its now routing locally.
Without getting too stuck into traceroutes – lets look international:
traceroute-nanog 216.66.84.250 traceroute to 216.66.84.250 (216.66.84.250), 30 hops max, 60 byte packets 1 ge-1-1-9-0.br-01-jnb.mtg.afpif.org (196.216.207.253) 21.574 ms 3.331 ms 3.821 ms 2 41.87.104.217 (41.87.104.217) 4.594 ms 3.843 ms 3.271 ms 3 41.217.213.154 (41.217.213.154) 184.253 ms 175.999 ms 176.560 ms 4 Gi9-0-0.mtz-201-access-3.mpls.seacomnet.com (41.217.213.194) 188.453 ms 184.564 ms 176.282 ms 5 41.217.213.21 (41.217.213.21) 187.913 ms 190.203 ms 184.012 ms 6 xe-7-0-0-0.par-gar-score-2-re1.interoute.net (84.233.162.169) 195.286 ms 197.172 ms 202.868 ms 7 ae0-0.par-gar-score-1-re0.interoute.net (212.23.42.25) 193.651 ms 196.539 ms 194.374 ms 8 ae1-0.lon-004-score-1-re0.interoute.net (212.23.42.22) 218.735 ms 204.191 ms 203.527 ms 9 10gigabitethernet1-1.core1.lon1.he.net (195.66.224.21) 198.666 ms 202.085 ms 216.741 ms 10 neology-ltd.10gigabitethernet5-4.core1.lon1.he.net (216.66.84.250) 202.736 ms 200.747 ms 212.251 ms
Last quick look at AS paths:
Now, lets take a quick look as speed – and YES – I don’t presume ‘speedtest.net’ to be a very effective REAL measure but it gets us a quick view:
To MWEB and Neology:
To Telkom:
And the ‘new Telkom’ speedtest:
and so on and so on….
So, what does this tell us:
- Afpif has its capacity from Seacom for the purpose of this conference
- Seacom has a peering point where they directly peer with MWEB – at a guess I’d say Teraco Isando
- Local transit is provided by Internet Solutions
- For some reason MTN is not routing this local even this is announcing this to Seacom. [This could be a number of things, but ‘feel’s like the typical MTN routing ‘funkiness’
- Seacom uses Interroute as their international upstream and they seem ‘reasonably’ connected at least.
- While we may have access to 2Gbps of Fiber it certainly does not appear that the local or international capacity is anywhere near that – not a problem per se – but not ideal
So, what about IPv6? Does it look the same or better:
Sadly – right now – not IPv6 at the conference, lets hope that gets fixed soon!
What am I really on about now:
Well, if more local participants were peering most of the issues highlighted here would not have been apparent. What’s needed:
- Seacom gets itself to JINX 😉
- MTN catches a wakeup and peers more freely
- Many of the local providers take another look at their routing engineering