ADSL speeds.

[twinscythe12332]

Hi
When doing your speed test, are you connecting to a server in the region that you are downloading from? International traffic has been up to crap since telkom started rolling out their connection upgrades.
You aren't the only ones, if that is any comfort.

http://mybroadband.co.za/news/adsl/94755-telkom-internet-adsl-problems.html

[Marq]

Here is my last tracerts -The first one is while downloading - the second is with no downloads.
The main difference or problem seems to be the link from durban to jhb. There is not much difference in the international speeds.
I am not sure how to read this but it sure looks like the differences are in the local areas.
The speedtest is from my machine to the local server - apparently its in ballito?

My speed went up to 8.6 for few days after I bitched but has been reduced to about 5 on average which is lower than before. Today its at 6.1. Seems there is no consistency.
As soon as as I download anything - say a you tube file - it goes down to 4.4
Download second you tube at same time, it goes down to 3.9 and so it will go down.
I am not sure why - but thought that the speed would be the same and it would just sort out the packets coming through (ie the new download itself may halve the time it takes to come through but the speed would be contestant.)

[Marq]

This blame game has been going on since the start of the computer age.
All the techies scratching heads and all the users frustrated and not knowing where to turn to.

It used to be ...its the software .....no its definitely hardware.
Hardware seems to have been replaced by .....network - that great nebulous pie in the sky area that has no home base or person in charge of any one section that we as users can turn to and actually get an answer.
Then suddenly some one somewhere along the chain flicks a switch or ticks a box and hey presto ...its all working again. Wasnt a problem in the first place - you users just dont know how to use the system.
Dont know how or why or where but its working and we can all individually or as a group along this chain receive credit and ra ra's.

Then another guy comes along and ticks the same box and ........so the cycle goes.

[Justloadit]

Could it be our new age techies don't actually understand how these systems work, after all 30% and over is considered a pass.

[irneb]

As soon as as I download anything - say a you tube file - it goes down to 4.4
Download second you tube at same time, it goes down to 3.9 and so it will go down.

As I described before, the speed test is itself a download (it simply downloads and then uploads some arb file/generated data and times it to find out how fast it went). Your line doesn't become faster in total because you're downloading more - it's like a water pipe: if you open one tap the water comes out at a certain speed (usually maximum the pipe can handle), when you open a second tap the 1st's speed drops (because the pipe's max hasn't increased by magic), if you open a 3rd both the 1st 2 drops in their speed (same non-magic pipe).

If you're downloading something else at the same time that you do a speed test, it means you're only getting a test showing how much is left of your download after sharing your line between 2 concurrent downloads. If you test while downloading 2 other stuff, then the test reveals how much is left when sharing your line between 3. These might not be equal to each other, it works more like a dif on a car - i.e. the one which downloads easier gets more speed. Thus it seems as if the test's download is easier than the youtube download and the 2nd download. The drop in the speedtest's speed due to the 1st youtube and the 2nd download shows that the drop is not proportional.

If you want to figure out exactly how much your line is getting in total you need to add the speed test's result to the speed the youtube is getting as well as the speed the 2nd download is getting. It's this total which you should gripe about, not the shared speed between downloads - that's always going to be less than the maximum. That's why you should stop all other downloads (including stuff like emails) while performing a speed test - it's the easiest way to try and get an accurate maximum speed test.

[irneb]

The difference between the tests on different days (if you didn't do them with various other downloads happening) would mean there's something else which is slowing the line down on that day. E.g. more people connected taday than yesterday means the main line shared between yourself and them is more congested. To use the water pipe analogy again: it would be as if your neighbour is filling his pool and you find that your shower head is now only dribbling.

Another thing that might be causing it is some restriction on the line, i.e. something went wrong somewhere. Again using the pipe - it's as if the pipe sprang a leak somewhere, or the plumbers are working on it and turned it off or rerouted your house (even temporarily) to use another pipe while they're doing something to the usual pipe.

[Marq]

Understood - thanks for explaining it to me again, now I have it.

If I was getting 4bar of pressure at the front of the house, I was expecting the reading still to be 4bar at the front of the house, even if I had all the taps on, which it would be. This is not the same instrument.

Now looking at those tracert results - these numbers represent the pressure outside the house from the reservoir . Are they the same as the speedtest results?

[irneb]

Now looking at those tracert results - these numbers represent the pressure outside the house from the reservoir . Are they the same as the speedtest results?

Those are measuring the "ping time" or "time-to-live" (TTL), otherwise called latency. Any internet connection can be tested for 2 things: baud-rate and latency. The baud-rate is what's measured by the speed test. To explain the difference:

Think of an internet connection like a telephone call:

• Latency is how long you wait after dialing before the other person picks up
• Baud-rate is how many words you can speak per time interval other after the person answered, and how fast they can understand such.

Those latency tests show how many milliseconds it took for each link in your internet connection to the destination server took to "answer the call". I'm not sure exactly how tracert formats its results, why there are 3 columns showing ms times I'm not too sure, and it's also not clear if those figures are running totals or latency per hop (I think it's a running total though). It also shows where the main issues are. E.g. here's mine (local IP's and domains redacted for security purposes):

C:\Users\Irne>tracert www.bbc.com

Tracing route to www-bbc-com.bbc.net.uk [212.58.246.94]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  #######
  2    <1 ms    <1 ms    <1 ms  #######
  3     1 ms     1 ms     1 ms  #######
  4     3 ms     2 ms     3 ms  #######
  5     2 ms     2 ms     2 ms  #######
  6     2 ms     2 ms     2 ms  ix-0-3-0-0.tcore1.JSO-Johannesburg.as6453.net [2
16.6.55.37]
  7   177 ms   178 ms   177 ms  41.206.178.2
  8   138 ms   138 ms   138 ms  195.219.214.25
  9   179 ms   178 ms   178 ms  80.231.158.1
 10   176 ms     *      175 ms  if-1-3.tcore1.SV8-Highbridge.as6453.net [80.231.
158.30]
 11   173 ms   174 ms     *     80.231.138.22
 12   176 ms   176 ms   176 ms  if-2-2.tcore1.L78-London.as6453.net [80.231.131.
 13   174 ms   174 ms     *     80.231.130.86
 14   173 ms   173 ms   173 ms  80.231.154.17
 15   177 ms   177 ms   177 ms  80.231.153.66
 16   175 ms   175 ms   175 ms  ae-70-70.csw2.Paris1.Level3.net [4.69.168.126]
 17   176 ms   176 ms   176 ms  ae-71-71.ebr1.Paris1.Level3.net [4.69.161.81]
 18   173 ms   173 ms   173 ms  4.69.143.101
 19   173 ms   174 ms   173 ms  4.69.153.118
 20   173 ms   173 ms   173 ms  4.69.166.5
 21   175 ms   175 ms   175 ms  212.113.14.222
 22     *        *        *     Request timed out.
 23     *        *        *     Request timed out.
 24   174 ms   174 ms   174 ms  ae0.er01.cwwtf.bbc.co.uk [132.185.254.93]
 25   175 ms   175 ms   175 ms  132.185.255.165
 26   178 ms   177 ms   177 ms  bbc-vip015.cwwtf.bbc.co.uk [212.58.246.94]

Trace complete.

This is using my office connection through a NeoTel 50MB/s fibre optic connection. Notice the latency jumps up significantly for the first hop to an international server (the Highbridge server). And most of the delays and timeouts happens overseas.

In your case though it seems the link from your local exchange to the jhb server takes longer to answer than my entire list of servers combined, for one of the tests, the other seemed to work quite well (anything below 100ms [1/10 of a second] is considered reasonable to good). Though your only timeout happened somewhere in BBC's own servers.

[irneb]

My baud-rate test though is a whole different matter. Local inside JHB:

Some other tests only show downloads of around 28MBits/s and uploads around 6MBits/s.

When running the test to a Cape Town server it fails on Latency. Seems there's a huge problem within SA.

And to an international connection:

[irneb]

If I was getting 4bar of pressure at the front of the house, I was expecting the reading still to be 4bar at the front of the house, even if I had all the taps on, which it would be. This is not the same instrument.

It is a similar instrument, because the speed test is as if you've placed the pressure gauge on one of the taps inside your house - not at the water meter at your gate.

[Dave A]

When running the test to a Cape Town server it fails on Latency. Seems there's a huge problem within SA.

Yep - and its name is Telkom

Marq - I'm actually stunned at your one set of results to the jhb.core server - I think the fastest I've ever got is around the 135 ms mark. Mostly it's closer to 185 ms.

Irneb - I'm seriously jealous.

[irneb]

Irneb - I'm seriously jealous.

Not the price though ... it's around R30k/month for the uncapped version. I suppose that's mitigated since it's shared between 70 people. We only recently (2 months ago) upgraded as we had some serious issues with the Teklom 10MB/s line ... it seems all over JHB Telkom's ADSL is simply stupidly slow - I've previously had dial-up speeds in Sandton, now in Bryanston it's not any better. Even my Vodacom phone gets a minimum of 4MBits/sec (sometimes nearing 10Mb/s) while every single ADSL I have seen here never gets more than 1MB/s (no matter what the "rating" is supposed to be).

That seems to indicate that Telkom IS the culprit: whenever your ISP needs to link through some TelHell line your speed WILL drop! As another sample: Last week we issued our Revit model to a local Online Document Management site (located inside JHB) for the other consultants on the project to use. The model is 200MBytes in size - it took 15 seconds to upload (near LAN speeds). But using an overseas site took 20 minutes for the same model. Not sure if that tcore bridge is one of telkom's or even one of those others.

Strangely today it seems perfectly fine: Testing to Telkom's own SAIX server (in Randburg) gets near max speeds today (upload is seriously awesome - similar to a WiFi LAN):


Linking to Telkom's Durban site gives similar speeds (ping's a bit less though):


To CT also:


Even to a site in London seems to work decently (ping's a bit high to my liking though):


It must've been something which was hogging the system yesterday?

That said: I know my sister (in Dubai) and a friend in the US has even better speeds than this from their residences. For less than a telkom 4MB/s line rental. We're still being screwed silly in SA, poor speed priced at an arm and a leg (and the rest of your family's limbs).

[AndyD]

Irneb, I've drawn up a list of things I'd like you to download for me please...

[irneb]

Irneb, I've drawn up a list of things I'd like you to download for me please...

Yeah! But then how to get it to CT inside this year?