I spend a lot of time using VNC to operate computers in the UK. So ping times (latency) directly affect my efficiency. 250ms translates to about 4 keystrokes a second whereas 180ms brings it up to 5.5. A 70ms drop in latency translates to a 40% improvement in productivity, something I'm always looking for.
So it is with a fair degree of interest that I look at my traceroute to London-based servers. It takes 15 hops to get to London, mostly because my packets make a point of always visiting New York before going to London. On the way back they do the same thing. With my little knowledge of South Africa's peering networks, this makes no sense. As far as I know, we're connected via two fibre networks:
I'm not routing expert, so I thought of trying to work out what the ping time should using another method. I used the VisualRoute demo to ping between London and New York. Ping time is 70ms. London to New York is 5500km. London to Cape Town is 9600km, therefore we should be able to go 70/5500*9600 which equals 120ms! That is less than half the latency I get on my unshaped ADSL account. What upsets me is that even though I'm paying R2000p/m for unshaped ADSL + bandwidth, I'm still getting a half-baked service.
It seems beyond doubt that we're getting screwed yet again, in new and interesting ways. Either our packets are being diverted to the US in Portugal or we're not using the fabled SAT3 cable at all. What is in question, yet again, is whether this can be explained as a malicious attempt by Telkom to cut costs at the expense of the Nation's users, or whether it is simply sheer incompetence. As usual, I doubt we'll ever find out.
Brendon
So it is with a fair degree of interest that I look at my traceroute to London-based servers. It takes 15 hops to get to London, mostly because my packets make a point of always visiting New York before going to London. On the way back they do the same thing. With my little knowledge of South Africa's peering networks, this makes no sense. As far as I know, we're connected via two fibre networks:
- SAT3 - West African fibre optic cable. Lands in Portugal
- SAFE - East African/Asia fibre optic cable - Lands in Malaysia
I'm not routing expert, so I thought of trying to work out what the ping time should using another method. I used the VisualRoute demo to ping between London and New York. Ping time is 70ms. London to New York is 5500km. London to Cape Town is 9600km, therefore we should be able to go 70/5500*9600 which equals 120ms! That is less than half the latency I get on my unshaped ADSL account. What upsets me is that even though I'm paying R2000p/m for unshaped ADSL + bandwidth, I'm still getting a half-baked service.
It seems beyond doubt that we're getting screwed yet again, in new and interesting ways. Either our packets are being diverted to the US in Portugal or we're not using the fabled SAT3 cable at all. What is in question, yet again, is whether this can be explained as a malicious attempt by Telkom to cut costs at the expense of the Nation's users, or whether it is simply sheer incompetence. As usual, I doubt we'll ever find out.
Brendon