International bandwidth throttled?

P

Peter7

Expert Member
Joined
Jun 13, 2005
Messages
1,005
Reaction score
1
Why is international connectivity so appalling at times? Why is there always a 300ms+ climb between the 3rd and 4th hop in any international tracert?
International connectivity isn't even stable, try watching streaming TV 24/7 and you'll see what I mean.

Traceroutes:

A) to Google:

3 23 ms 22 ms 23 ms wblv-ip-er-1-fe-11-1-1-1.telkom-ipnet.co.za [196.43.11.214]
4 361 ms 352 ms 355 ms lon-ip-dir-telecity-pos-6-0.telkom-ipnet.co.za [196.43.9.50]

B) to CNN:

3 24 ms 23 ms 22 ms wblv-ip-er-1-fe-11-1-0-1.telkom-ipnet.co.za [196.43.10.138]
4 360 ms 366 ms 357 ms ash-ip-dir-equinix-pos-5-1.telkom-ipnet.co.za [196.43.9.106]

There's virtually no transmission time discrepancy with tracerts between UK and the US?! Only good 'ol SA to the rest of the world. Is TELKOM throttling majorly on the SAT3 link and other links or something?

Can someone shed some light on this?
:confused:
 
Last edited:
They're not throttling bandwidth, only artificially adding latency so they can pimp their "unshaped" accounts at a higher price.
 
Perdition said:
They're not throttling bandwidth, only artificially adding latency so they can pimp their "unshaped" accounts at a higher price.
Click to expand...
Hmmm how would they do that? This is a discussion that I know many of use have had hundreds of times and there are many threads on the forums about it, Just wondering how they would do that as that is about the best explaination for it so far? I know that if somebody does a ping to SA from the USA they also get the large ping, yet they can Ping Russia in half the time and with 10 more hops than to SA?
 
There is always a big hop in ping times as one crosses the ocean to the other side of the world. The speed of light is an issue over the distance involved and boosting the signal etc. probably also has an effect. Ping times over satellite is a lot worse, since the distance the signal has to travel is so much further. I do not think you can blame Telkom for this one. ;)
 
gkm said:
There is always a big hop in ping times as one crosses the ocean to the other side of the world. The speed of light is an issue over the distance involved and boosting the signal etc. probably also has an effect. Ping times over satellite is a lot worse, since the distance the signal has to travel is so much further. I do not think you can blame Telkom for this one. ;)
Click to expand...
I'm afriad you can blame telkome with that one. There is a thread here where we did all the calculations, including over heads added extra for the distance for arugments sake and doubled it all.. Its still a lot less than what we are getting and as Peter said and I said, Ping times from the UK to to USA, or Russia to the USA are almost half what we are getting, its the same distance and some times longer... SO its either telkom or the Cables that are bad
 
The cables can't be bad. Unlike copper if fibre is bad you get zero connection ;) Consider this, if Telkom are able to offer an unshaped account that consistently gives you pings of 250ms to the UK and the shaped accounts run over EXACTLY the same network and fibre, then where does the extra latency come from? It's definitely artificially introduced as SAT-3 traffic is nowhere near the capacity that would affect latency.
 
Analysis

There's definitely a "problem" or bottleneck somewhere. I work for a US company and I have remote access to several machines in various colo facilities over there. When I do trace routes from these US machines, to the UK (bbc.co.uk), there's only a 70ms difference across the ocean. However, when I do trace routes to SA (saix.net) there's that 230ms discrepancy that we're talking about.

Conclusion: This 230ms discrepancy seems to only occur for traffic entering or leaving South Africa. This seems to indicate that there is a bottleneck at the routers for traffic to/from SA. Are these routers just archaic (OLD), and Telkom doesn't want to fork out any money to upgrade them to match the equipment used between the US and the UK? (If this is the case, I hope they don't use their infamous excuse: You can't compare us to a developed nation.)

Try access http://www.dnsstuff.com/ and have their server which is in Washington, Washington DC do a traceroute and you'll see the same results. (For some reason it's not resolving from my local ADSL connection.) When you analyze the traceroutes below notice the following (note I've highlighted the best response times and COUNTRY codes of the routers in the trace):

To the UK:
-----------
- The 70ms increase over the ocean, from the US to the UK, on hop 6.

To South Africa:
----------------
- The 230ms increase from Europe to South Africa, on hop 6.
- If the country codes for the routers are correct, then there's no ocean even involved here! This seems to indicate that latency in question occurs upon traffic entering and leaving South Africa (through Telkom's network)


To the UK: Tracert to bbc.co.uk
Generated by www.DNSstuff.com at 15:09:30 GMT on 20 Jul 2005.
Hop T1 T2 T3 Best Graph IP Hostname Dist TTL Ctry
1 0 0 1 0.5 ms 66.36.240.2 AS14361HOPONE-DCA c-vl102-d1.acc.dca2.hopone.net. 255 US
2 0 0 1 0.5 ms [+0ms] 66.36.224.227 AS0IANA-RSVD-0 ge5-0.core2.dca2.hopone.net. 254 US
3 1 2 2 1.8 ms [+1ms] 66.36.224.34 AS0IANA-RSVD-0 ge2-0-241.core1.iad1.hopone.net. 253 US
4 2 2 2 2.0 ms [+0ms] 206.223.115.143 AS0IANA-RSVD-0 eqix-gw0.rt0.eqash.bbc.co.uk. 252 US
5 8 7 6 6.5 ms [+4ms] 212.58.255.114 AS9156BBC-US sw0.thdo.bbc.co.uk. 252 UK
6 80 80 80 78 ms [+72ms] 212.58.255.33 AS9156BBC-US rt0-pos3-0.111ny.bbc.co.uk. 250 UK
7 79 78 78 78 ms [+0ms] 212.58.255.214 AS9156BBC-US unknown.co.uk 252 UK
8 100 * 84 84 ms [+5ms] 212.58.224.131 AS2818BBC[Reached Destination]rdirwww-vip.thdo.bbc.co.uk. 122 UK


To SA: Tracert to www.saix.net
(Generated by www.DNSstuff.com at 15:13:50 GMT on 20 Jul 2005.)

Hop T1 T2 T3 Best Graph IP Hostname TTL Ctry
1 0 0 0 0.5 ms 66.36.240.2 AS14361 HOPONE-DCA c-vl102-d1.acc.dca2.hopone.net. 255 US
2 0 1 0 0.5 ms [+0ms] 66.36.224.243 AS0IANA-RSVD-0 ge4-0.core2.dca2.hopone.net. 254 US
3 0 0 0 0.5 ms [+0ms] 66.36.224.226 AS0IANA-RSVD-0 ge4-0.core1.dca2.hopone.net. 254 US
4 5 6 5 5.8 ms [+5ms] 66.36.224.174 AS0IANA-RSVD-0 unknown.hopone.net 252 US
5 5 6 5 5.8 ms [+0ms] 195.2.5.86 AS1273CW telecomso.nyk.cw.net. 252 EU
6 249 234 234 234 ms [+228ms] 196.43.11.30 AS5713AFRINIC-ASNBLOCK-5713 wblv-ip-essr-1-atm-2-0-0-2.telkom-ipnet.co.za. 247 ZA
7 * * 238 238 ms [+3ms] 196.25.1.200 AS5713AFRINIC-ASNBLOCK-5713 [Reached Destination]wblv-ip-www-1.telkom-ipnet.co.za. 55 ZA


Like I've said earlier this seems to indicate that latency occurs on the Telkom routers when traffic enters/leaves South Africa. This same latency is not seen for the trans-oceanic link between the UK and the US. So what's up here? Are they using old equipment b/c they don't want to spend money upgrading?
 
Last edited:
I think telkoms backend is still running on multilinked 56k modems whaha ;-)

*sigh*
 
Its definately possible though to get around 150-180ms to the UK over the SAT3. Try run a traceroute from the UK to picasso.wcape.school.za or lordcow.phy.uct.ac.za... UCT has a nice dedicated sat3 sector by the looks of things... You may also want to trace an international IP (ie: clarity.jolt.co.uk) from http://services.truteq.com/cgi-bin/nph-traceroute which is hosted at POSIX in JHB... also ~170ms pings to the UK.. so its clearly Telkoms imposed shapes on the latency that makes the whole issue, especially on the wblv saix router by the looks of things, which is where most international ADSL routing goes if I'm not mistaken?
 
nope it goes from cpt to jhb to namibia and there is the big cabling system in and out, check on the worldmap somewhere. Capetown users will allways have higher ping as we need to route to jhb first.

using that I get to google.com 330ms and to cs.jolt.co.uk I get 210ms.

Ok also note, when users join south-african CS servers they also sit with 300ms odd, similar to SA to UK over adsl. Suppose we call this an issue with ZA international, fair enough.. but that screws up your point regarding pinging from international side to local side being lower than vica versa.
 
Last edited:
From my boss's 1Mbps Bellsouth DSL account in the US, the tracert to my SA machine is consistent with the results I posted earlier to saix.net:
13 30 ms 30 ms 40 ms ny-ip-dir-globalc.telkom-ipnet.co.za [196.43.9.250]
14 341 ms 350 ms 340 ms rrba-ip-lir-1-pos-6-0.telkom-ipnet.co.za [196.43.9.113]

Here's the result from my machine to his in the US:
3 22 ms 22 ms 21 ms wblv-ip-er-1-fe-11-1-1-1.telkom-ipnet.co.za [196.43.11.214]
4 341 ms 340 ms 339 ms ash-ip-dir-equinix-pos-5-1.telkom-ipnet.co.za [196.43.9.106]

US -> ZA: 300+ms climb
ZA -> US: 300+ms climb

I'm not sure what the deal is with the posix lookup being much lower for ZA to international (from the resolved domain names the traceroute doesn't appear to go through any Telkom router) From my connection traceroutes to those 2 hosts, avg 340ms.

Here's the latest topology diagram from TELKOM/SAIX: http://saix.net/diagrams/ipnet_jul05.pdf
 
Last edited:
some tracerts from my uk dsl connection for comparisons..

Jolt a game server hosted on the uk jolt network there very fast good for checking latency.

Tracing route to clarity.jolt.co.uk [195.149.21.11]
over a maximum of 30 hops:

1 23 ms 25 ms 23 ms gauss-dsl.zen.net.uk [62.3.82.18]
2 24 ms 25 ms 22 ms deleuze-ge-0-1-0-1.hq.zen.net.uk [62.3.80.161]
3 32 ms 29 ms 32 ms suarez-so-0-0-0-0.te.zen.net.uk [62.3.80.62]
4 30 ms 29 ms 30 ms lon1-9.nildram.net [195.66.224.59]
5 29 ms 29 ms 30 ms jolt-gw.nildram.net [195.149.20.126]
6 31 ms 29 ms 32 ms clarity.jolt.co.uk [195.149.21.11]

Trace complete.

google.com

Tracing route to www.l.google.com [66.102.9.99]
over a maximum of 30 hops:

1 23 ms 22 ms 24 ms gauss-dsl.zen.net.uk [62.3.82.18]
2 22 ms 24 ms 22 ms deleuze-ge-0-1-0-1.hq.zen.net.uk [62.3.80.161]
3 30 ms 32 ms 30 ms suarez-so-0-0-0-0.te.zen.net.uk [62.3.80.62]
4 28 ms 29 ms 29 ms ge3-0.pr1.lhr1.uk.above.net [195.66.224.76]
5 29 ms 29 ms 28 ms pos3-0.mpr2.lhr1.uk.above.net [208.184.231.73]
6 49 ms 47 ms 147 ms 213-152-230-99.google.net [213.152.230.99]
7 49 ms 48 ms 49 ms 216.239.46.173
8 62 ms 63 ms 60 ms 216.239.49.254
9 59 ms 58 ms 60 ms 64.233.174.42
10 59 ms 61 ms 61 ms 64.233.174.14
11 59 ms 59 ms 59 ms 66.102.9.99

Trace complete.

tracert too picasso.wcape.school.za

Tracing route to picasso.wcape.school.za [196.14.20.12]
over a maximum of 30 hops:

1 23 ms 22 ms 21 ms gauss-dsl.zen.net.uk [62.3.82.18]
2 21 ms 22 ms 22 ms deleuze-ge-0-1-0-1.hq.zen.net.uk [62.3.80.161]
3 29 ms 29 ms 29 ms suarez-so-0-0-0-0.te.zen.net.uk [62.3.80.62]
4 30 ms 29 ms 30 ms 195.66.224.198
5 29 ms 29 ms 29 ms 168.209.246.4
6 292 ms 292 ms 295 ms 168.209.0.89
7 294 ms 293 ms 291 ms 168.209.2.8
8 196 ms 198 ms 195 ms 196.23.188.147
9 198 ms 204 ms 197 ms 168.209.2.135
10 310 ms 308 ms 317 ms 168.209.29.214
11 217 ms 214 ms 213 ms picasso.wcape.school.za [196.14.20.12]

Trace complete.

tracert to lordcow.phy.uct.ac.za

Tracing route to lordcow.phy.uct.ac.za [137.158.36.179]
over a maximum of 30 hops:

1 24 ms 23 ms 24 ms gauss-dsl.zen.net.uk [62.3.82.18]
2 22 ms 23 ms 25 ms deleuze-ge-0-1-0-1.hq.zen.net.uk [62.3.80.161]
3 29 ms 29 ms 34 ms suarez-so-0-0-0-0.te.zen.net.uk [62.3.80.62]
4 30 ms 28 ms 29 ms cr02.ldn01.pccwbtn.net [195.66.224.167]
5 281 ms 280 ms 290 ms wblv-ip-lir-1-pos-6-0.telkom-ipnet.co.za [196.43
.9.49]
6 278 ms 278 ms 274 ms tenet-gw.wc.saix.net [196.25.55.66]
7 428 ms 387 ms 358 ms int.uct.uni.net.za [155.232.50.245]
8 * * * Request timed out.

2 last interesting ones

tracert to japan!

Tracing route to home.hi-ho.ne.jp [202.224.157.80]
over a maximum of 30 hops:

1 21 ms 22 ms 23 ms gauss-dsl.zen.net.uk [62.3.82.18]
2 22 ms 22 ms 23 ms deleuze-ge-0-1-0-1.hq.zen.net.uk [62.3.80.161]
3 22 ms 22 ms 22 ms lotze-ge-0-0-0-0.hq.zen.net.uk [62.3.80.57]
4 23 ms 26 ms 26 ms bolzano-ge-0-0-3-0.wh.zen.net.uk [62.3.80.53]
5 22 ms 22 ms 22 ms 195.16.169.89
6 23 ms 24 ms 22 ms so-10-0.hsa1.Manchesteruk1.Level3.net [4.68.113.
97]
7 173 ms 172 ms 175 ms ae-0-0.bbr2.SanJose1.Level3.net [64.159.1.130]
8 171 ms 170 ms 172 ms ae-14-51.car4.SanJose1.Level3.net [4.68.123.14]

9 171 ms 172 ms 171 ms unknown.Level3.net [64.152.102.62]
10 279 ms 279 ms 279 ms po1-0.cr2.nrt1.asianetcom.net [202.147.0.194]
11 280 ms 280 ms 279 ms gi5-2.gw1.nrt2.asianetcom.net [202.147.0.166]
12 279 ms 280 ms 278 ms Matsu****a.asianetcom.net [203.192.149.238]
13 292 ms 295 ms 293 ms 202.224.130.139
14 300 ms 298 ms 282 ms g02.c3550-12g2.dojima.pana.net [202.224.128.202]

15 293 ms 300 ms 297 ms home.hi-ho.ne.jp [202.224.157.80]

Trace complete.

last tracert too moscow russia

Tracing route to cs.playground.ru [212.42.38.194]
over a maximum of 30 hops:

1 22 ms 23 ms 24 ms gauss-dsl.zen.net.uk [62.3.82.18]
2 26 ms 26 ms 25 ms deleuze-ge-0-1-0-1.hq.zen.net.uk [62.3.80.161]
3 21 ms 22 ms 23 ms lotze-ge-0-0-0-0.hq.zen.net.uk [62.3.80.57]
4 30 ms 23 ms 28 ms bolzano-ge-0-0-3-0.wh.zen.net.uk [62.3.80.53]
5 22 ms 24 ms 22 ms 195.16.169.89
6 24 ms 25 ms 24 ms so-4-1-0.mp1.Manchesteruk1.Level3.net [4.68.113.
101]
7 30 ms 31 ms 33 ms so-1-1-0.bbr1.London1.Level3.net [4.68.128.73]
8 33 ms 32 ms 32 ms ge-1-1.core2.London1.Level3.net [212.187.131.134
]
9 31 ms 31 ms 34 ms CABLE-WIRE.c2.Level3.net [212.73.240.170]
10 33 ms 31 ms 29 ms so-5-2-0-dcr2.tsd.cw.net [195.2.10.134]
11 30 ms 29 ms 30 ms as0-dcr1.tsd.cw.net [195.2.10.165]
12 37 ms 39 ms 37 ms so-0-0-0-dcr2.amd.cw.net [195.2.10.145]
13 49 ms 50 ms 51 ms so-2-0-0-ycr1.cpi.cw.net [195.2.10.182]
14 60 ms 58 ms 56 ms so-1-0-0-ycr1.skt.cw.net [208.173.216.26]
15 58 ms 57 ms 60 ms ge-0-3-0-zar1.skt.cw.net [166.63.220.131]
16 77 ms 75 ms 79 ms goldentel6.skt.cw.net [166.63.220.98]
17 81 ms 83 ms 79 ms cisco13.Moscow.gldn.net [194.186.157.185]
18 77 ms 78 ms 79 ms cat02.Moscow.gldn.net [194.186.157.237]
19 88 ms 82 ms 82 ms Ropnet-gw.Moscow.gldn.net [194.186.0.90]
20 86 ms 180 ms 84 ms M9-0.ropnet.ru [212.42.36.254]
21 83 ms 85 ms 86 ms M47-100.ropnet.ru [212.42.36.158]
22 83 ms 84 ms 84 ms www.playground.ru [212.42.38.194]

Trace complete.
 
Thanks BlackMesa for your UK traceroutes. The 2 traceroutes to ZA are indicative of the issue we're discussing on this thread, and just reinforces the argument that the problem is on ZA's side or rather more accurately TELKOM. The 280ms+ increase on these 2 traceroutes is consistent with the the traceroutes I've done from various locations in the US. From an analytical standpoint - just reviewing the facts, traffic from ZA to UK/US shows the 230+ms increase as it leaves ZA and again this 230+ms increase when entering ZA. WHAT'S UP TELKOM???
 
Everything said there has been said for years now, do you think realisticly we will see 1% of that come true? I think not..

Lets hope for only a few of that happening, they all mention, "Suggest to Telkom" which means, lets hope they dont feel like making us bend.
 
An interesting (original) discussion.

The best I've seen from a local ISPs core network, with FAT pipes into SAIX, is 170ms overseas (to London I think). That still leaves a 100ms difference between US-UK & SA-UK.

Not defending Telkom, but there are some physical/technical factors to consider ...

1. Plain & simple the distance is less. Undersea cables between US-UK are probably a straight run. The SAT3 cable hugs the west African coast, which means its distance is far longer than the crow flies (which is further anyway).

2. Due to economics I would think the quality & sophistication of the fiber/repeaters/technology used on transatlantic links is probably the best avaliable while SAT3 is probably more cost effective/ruggered.

3. Transalantic links are basically point to point, while SAT3 has many hop off points along the African coast. At these points there must signal splitting equipment that probably also adds latency.

Not sure if the above covers the 100ms differential seen between US-UK & SA-UK?
 
All this quabble still does not explain why in 2001, I could play BZ Flag against people in Denmark at under 150ms lag (Admittedly at night or on weekends), office hours it'd be erratic jumping up to 220ms at times.

Then they introduced shaping - now i'll never get less than 350, no matter what.

So - by simple deduction, yes - even with that massive distance, you must be able to reach UK in uder 150ms, and, shaping caused the delay.

I'm trying to get hold of somebody further up the SAT3 line, say NIGERIA. for further tests, but having difficulty finding anybody.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X