Best ADSL ISP network in South Africa

Too bad WA shape, throttle and have rolling windows ;)

Plugg ftw :D
 
Irrelevant title, only 5 ISP's rated
It is fairly relevant, the ISPs listed run their own networks. Title does say "ADSL ISP network" and not "ADSL ISP" only.

Although, I personally would have liked to see weighted averages instead of just average. Ie. Average linked to the amount of tests per ISP network, since I am sure cybersmart and WA gets a lot less tests from their networks than IS and Telkom. So 2 more columns, 1 for amount of tests from network, and then another with the weighted average, with amount of tests as the weight.
 
Hold on am I missing something...10mb lines, running at 4mb speeds?
 
So the average 10mb ADSL line user in SA is using a 4mb account? Clearly pricing isn't quite where it's supposed to be yet. Looking forward to a good laugh when ISPs announce their 20mb/40mb uncapped pricing.
 
Does not make sense - MWEB has the best latency by a long shot and thus should have the best speed.

The suspicion is that the ISP who is hosting the speedtest server has some dpi or congestion somewhere?

BTW: Speedtest latency measurement seems way off. Here are my figures to speedtest.mybroadband.co.za:
mybroadband_latency.png
 
Irrelevant title, only 5 ISP's rated
As Tinuva highlighted, most of the smaller ISPs use Internet Solutions or Telkom for their networks. The network performance for all these ISPs will therefore be the same. Since we use IP addresses to ensure our information is correct, the smaller ISPs cannot be ranked...

BTW: We are currently re-doing the system to give even more fine grained information, but even with the new system reseller ISPs will not be listed (just not possible).
 
Does not make sense - MWEB has the best latency by a long shot and thus should have the best speed.

The suspicion is that the ISP who is hosting the speedtest server has some dpi or congestion somewhere?

BTW: Speedtest latency measurement seems way off. Here are my figures to speedtest.mybroadband.co.za:
View attachment 35594
iBurst is in the same DC as the speed test, so not really fair ;)

We are considering moving the Speedtest server to Teraco to make it more neutral.
 
Mweb has the most stable network (very little down time) the lowest latency, the least packet loss and the best uncapped policy out of all the ISP's! Just my 2c
 
Does not make sense - MWEB has the best latency by a long shot and thus should have the best speed.

The suspicion is that the ISP who is hosting the speedtest server has some dpi or congestion somewhere?

BTW: Speedtest latency measurement seems way off. Here are my figures to speedtest.mybroadband.co.za:
View attachment 35594
Latency can affect throughput, on UDP at least. But TCP is not as badly affected by latency, and as such latency is not the single dominator for throughput as you make it out to be.
 
Does not make sense - MWEB has the best latency by a long shot and thus should have the best speed.

The suspicion is that the ISP who is hosting the speedtest server has some dpi or congestion somewhere?

BTW: Speedtest latency measurement seems way off. Here are my figures to speedtest.mybroadband.co.za:
View attachment 35594

Holy crap sticks Batman, 2.5ms! Are you next door to their servers?

Here is our offices ping to MyBB using OpenWeb Business Gold (IS network):

Screenshot%20at%202013-01-16%2011%3A30%3A08.png
 
It is fairly relevant, the ISPs listed run their own networks. Title does say "ADSL ISP network" and not "ADSL ISP" only.

Although, I personally would have liked to see weighted averages instead of just average. Ie. Average linked to the amount of tests per ISP network, since I am sure cybersmart and WA gets a lot less tests from their networks than IS and Telkom. So 2 more columns, 1 for amount of tests from network, and then another with the weighted average, with amount of tests as the weight.

This has got to be one of my biggest pet peeves... If you can quantify (count) something, the correct word to use is "number". For example, "Average linked to the number of tests..." Amount is only used for mass or uncountable nouns.

See here for more info > http://grammarist.com/usage/amount-number/

I know this is irrelevant to the topic, but I just couldn't help myself!
 
This has got to be one of my biggest pet peeves... If you can quantify (count) something, the correct word to use is "number". For example, "Average linked to the number of tests..." Amount is only used for mass or uncountable nouns.

See here for more info > http://grammarist.com/usage/amount-number/

I know this is irrelevant to the topic, but I just couldn't help myself!
Thank you, but really? English is my second language, you can not really blame me for not knowing every single grammar mistake, especially since I am not a writer or anything even remotely related to that.
 
Latency can affect throughput, on UDP at least. But TCP is not as badly affected by latency, and as such latency is not the single dominator for throughput as you make it out to be.

WTF??? TCP is heavily affected by latency, a hell of a lot more than UDP. UDP it all keeps flowing, TCP requires an ACK before the next set of packets gets sent, therefore you sit with idle line speed while the sender waits for the ACK. UDP, it just keeps on flowing (unless higher level in the stack has some control function to do checks of course). Remember TCP has a mximum window size, so no matter how fast your connection is TCP will be a limiting factor when latency comes in to play.

Round trip latency CP Throughput

0ms 93.5 Mbps

30ms 16.2 Mbps

60ms 8.07 Mbps

90ms 5.32 Mbps

Hope this clears a few things up!
 
Does WebAfrica still only authenticate on CINX or have they extended their service to JINX ?
If not, it means that a WebAfrica ADSL client in Joburg still needs to make a Cape Town round-trip, causing a load on the Latency - and that is the main reason I don't use them as my Primary ISP anymore.
 
Last edited:
Top
Sign up to the MyBroadband newsletter