Telkom Internet Capped/Uncapped User Feedback (Pt2)

Got back Home from university on the 9 June. I been home for almost two weeks now and the internet is horrible. I must confess I play alot of online games when I am at home, especially league of legends. When my internet is working perfectly the ping to league of legends server (Frankfurt,Germany) is normally between 200ms and 250ms which is acceptable for me. Must say during the december 2013 holiday when i was using the mweb trail 4mb uncapped account (my line is only 2mb) got a stable 178ms - 190ms. Lately my ping is so unstable I dont even want to play league any more. During a game it will jump between 250ms and 500ms constantly, if i am lucky it would only go as high as 350ms. This is happening during the day (10:00 - 16:00) . When the clock strikes 17:00 I can forget about playing league until like 23:30 at night because my ping is anywhere between 300ms and 1100ms during this time.

The worst part is this unstable and high ping has been happening every f**king day since i got home. Needless to say we have a 20gig capped telkom account and there is still 8gigs left for the month. A few day ago i found this article (http://mybroadband.co.za/news/adsl/82251-how-to-test-for-adsl-exchange-congestion.html) and in the comments i read about pingplotter. I downloaded and installed the pro version. The program has been running since 18 june on my computer.

Just to be clear: Pingplotter is doing a traceroute to telkom.co.za
1,login.router,10.0.0.2
2,ti-224-96-01.telkomadsl.co.za,105.224.96.1
3,ti-226-0-22.telkomadsl.co.za,105.226.0.22
4,ti-226-0-57.telkomadsl.co.za,105.226.0.57
5, -------------- ,165.165.214.213
6, -------------- ,196.43.39.166
7,rrba-ip-bssr-1-ge-2-48.telkom-ipnet.co.za,196.43.23.6
8,nbsc-ip-bssr-2-atm-5-0-0-1.telkom-ipnet.co.za,196.43.23.34
9,telkom.co.za,196.43.22.222

Pingplotter is doing the traceroute to telkom.co.za every 15 seconds. The nice thing i found about PingPlotter is that you can export the result to a text file. This is one of the traceroute results : "6/18/2014 2:21:16 PM",0,11,17,13,13,34,40,42,35
At the moment i have over 17000 of these traceroute results.

So i wrote a small java program to help me analyze this data. As you can see each traceroute results has a date and time. I created a 2D array with the following properties: day,month,year,hour,minute,second,am_pm,time_interval and the traceroute results. Here is and example how it looks: (6 21 2014 3 24 48 AM 3 0 10 13 12 11 34 33 34 37 ) the 3 after AM tells us this traceroute was done between 03:00:00 and 03:59:59. and after that is the results for the traceroute (hop 1 = 0ms ... hop 9 = 37ms) And from this i can do different queries

One of the queries that i have done on the data: I calculate the average ping to telkom.co.za for each time interval from 17000 and still growing traceroute results. I will post the results when i am doing implementing the variance, median, and standard deviation for each hop in each time interval. So that they can see how unstable the ping is.

They want proof i will give them proof.
 
Results

Got some interesting results.
Link:https://www.dropbox.com/s/dr1ncul1tdq2rev/Hop9IntervalTable.csv

The above table only show the results for hop number 9. Hop number 9 is the last hop in the tracerout.

Turns out the best time to play online games is between 04:00:00 and 07:59:59 in the morning. This is when the lowest ping absolute deviation is closes to zero.

The worst time to play is between 17:00:00 and 23:59:59. This when ADL Ping is above 150 signaling that ping will be
all overt the place.
 
Brennan123 said:
Got some interesting results.
Link:https://www.dropbox.com/s/dr1ncul1tdq2rev/Hop9IntervalTable.csv

The above table only show the results for hop number 9. Hop number 9 is the last hop in the tracerout.

Turns out the best time to play online games is between 04:00:00 and 07:59:59 in the morning. This is when the lowest ping absolute deviation is closes to zero.

The worst time to play is between 17:00:00 and 23:59:59. This when ADL Ping is above 150 signaling that ping will be
all overt the place.
Click to expand...

In other words play when nobody else is there to play with.
 
Brennan123 said:
Got back Home from university on the 9 June. I been home for almost two weeks now and the internet is horrible. I must confess I play alot of online games when I am at home, especially league of legends. When my internet is working perfectly the ping to league of legends server (Frankfurt,Germany) is normally between 200ms and 250ms which is acceptable for me.

[...]

The worst part is this unstable and high ping has been happening every f**king day since i got home. Needless to say we have a 20gig capped telkom account and there is still 8gigs left for the month.
Click to expand...

But, this thread is about uncapped, and our capped products have a much better experience during high demand. You seem to be in the wrong thread.

A few day ago i found this article (http://mybroadband.co.za/news/adsl/82251-how-to-test-for-adsl-exchange-congestion.html)
Click to expand...

Firstly, again, you seem to be in the wrong thread. This is not the "bitch about everything Telkom" thread. It is not the "DSLAM congestion" thread. It is a thread specifically about Telkom Internet uncapped, where feedback from users is welcome, and any new concerns can be highlighted, and which has (via me) some visibility from the very small team that develops and supports the Telkom Internet products (from the technical side).

Secondly, all that article proves is how clueless some of MyBroadband's technical go-to people are.

Testing for DSLAM congestion is *very* simple:
1)Connect to your modem with a network cable. Disconnect any other PCs connected by network cable except the one used for this test.
2)Disable WiFi on your modem
3)Create a new PPPoE connection on your PC with the username 'guest@telkomadsl' with the password 'guest'.
4)Point your browser to http://www.saix.net, click on the speedtest button, and do a speedtest to the closest site. Screenshot the result (and copy/paste the text below as well).
5)For extra credit, traceroute to e.g. capetown.spdtst.saix.net

If you find latency higher than 50ms, log a line fault online, with the description that the DSLAM is congested, and please find a different thread to complain in when your fault is closed (like mine was). Maybe you can send your fault number to TelkomZA, I don't know if Charl and the rest of the team have more leverage on getting DSLAM backhauls upgraded.


and in the comments i read about pingplotter. I downloaded and installed the pro version. The program has been running since 18 june on my computer.
Click to expand...

Your initiative may have been easier with mtr (depending on your skills on Linux or mingw32).

Just to be clear: Pingplotter is doing a traceroute to telkom.co.za
1,login.router,10.0.0.2
2,ti-224-96-01.telkomadsl.co.za,105.224.96.1
3,ti-226-0-22.telkomadsl.co.za,105.226.0.22
4,ti-226-0-57.telkomadsl.co.za,105.226.0.57
5, -------------- ,165.165.214.213
6, -------------- ,196.43.39.166
7,rrba-ip-bssr-1-ge-2-48.telkom-ipnet.co.za,196.43.23.6
8,nbsc-ip-bssr-2-atm-5-0-0-1.telkom-ipnet.co.za,196.43.23.34
9,telkom.co.za,196.43.22.222
Click to expand...
If this is showing high latency, your problem is almost guaranteed to be DSLAM congestion. If you had actually included the output from just plain tracert at a time when you are experiencing high latency, we could have told you for sure (or at least with a high level of confidence).

Pingplotter is doing the traceroute to telkom.co.za every 15 seconds. The nice thing i found about PingPlotter is that you can export the result to a text file. This is one of the traceroute results : "6/18/2014 2:21:16 PM",0,11,17,13,13,34,40,42,35
At the moment i have over 17000 of these traceroute results.

So i wrote a small java program to help me analyze this data. As you can see each traceroute results has a date and time. I created a 2D array with the following properties: day,month,year,hour,minute,second,am_pm,time_interval and the traceroute results. Here is and example how it looks: (6 21 2014 3 24 48 AM 3 0 10 13 12 11 34 33 34 37 ) the 3 after AM tells us this traceroute was done between 03:00:00 and 03:59:59. and after that is the results for the traceroute (hop 1 = 0ms ... hop 9 = 37ms) And from this i can do different queries

One of the queries that i have done on the data: I calculate the average ping to telkom.co.za for each time interval from 17000 and still growing traceroute results. I will post the results when i am doing implementing the variance, median, and standard deviation for each hop in each time interval. So that they can see how unstable the ping is.

They want proof i will give them proof.[/QUOTE]

Did someone ask for proof? Really, no-one you will be able to talk to at the call centre is going to 1)Understand what you have done, and why that has any relevance to the script they are reading from, 2)Be able to do anything about it.

However, if you had just provided the traceroute (or winmtr, I haven't used pingplotter, winmtr is enough to identify any problem) output, we could have known where the problem lies (90% chance it is DSLAM congestion if your data to telkom.co.za shows high latency).
 
Just got managed on 238.87GB I thought you could do about 350GB to 400GB on a 4mb uncapped account without being managed.

Or is that in the old days....

So what are the thresholds for Telkom's uncapped products?

Chocked down to about 10% of the line speed.
 
Last edited:
3579030683.png


After downloading BF4 since 11:30 PM yesterday.
 
Dade Murphy said:
Apparently it seems to depend on how much of your downloads are between 12am-6am.
Click to expand...

Will have to test that theory...

For some reason userstats.adsl.saix.net show I did 37GB in 24hrs... that just does not sound right.
 
KingLv said:
Just got managed on 238.87GB I thought you could do about 350GB to 400GB on a 4mb uncapped account without being managed.

Or is that in the old days....
Click to expand...

That is the old days, when you were paying R120/month more, but usage when usage between midnight and 06h00 was taken into account in fair usage enforcement.

If you were not incurring any usage during nightsurfer time, you get a bit less usage for a lot less money, if you were scheduling downloads for nightsurfer time, you get a lot more usage for a lot less money, if you were running at constant speeds 24 hours a day, you get about the same usage for much less.

It is unfortunate that this wasn't explained more explicitly at the time, but we hope to have some information on tracker at some point.

Why are we doing this? To move less valuable traffic (e.g. p2p) away from times when real-time traffic (e.g. streaming, video-conferencing) is more important, but (ideally) without making p2p unuseable during peak times (which is one of the other options, with its own problems).
 
MickeyD said:
Do you have to start the session after 0h00 for it not to count towards FUP? Or can you start the download at (say) 22h00 and it runs past midnight?
Click to expand...

As far as I know, any new 'flow' (in the case of tcp, a connection, in the case of non-tcp packets that share source,destination,source-port, destination port) that is started after midnight and finishes before 06h00 will not be counted for FUP purposes. Existing flows can possibly be accounted for a few minutes. Also, flows existing before 06h00 may not be counted for a few minutes.

My scheduling is set to apply different limits from 00h05 and pause all running downloads and drop limits around 05h55. However, usually my downloads are started by rss feed around 04h00, and finish by 6. In the event a feed publishes a file late that I want sooner, I bite the bullet and download at about half line speed when I remember to check the feed (usually an hour or so before I want it).

But, at present due to limited time, I can't really make effective use of more than about 50GB/month (with a lot of other weird traffic). At times, youtube is serving us 720p videos by 06h30, or at others I have work to do until midnight, so I also schedule the downloads to not affect such important activities.
 
Last edited:
ranger said:
That is the old days, when you were paying R120/month more, but usage when usage between midnight and 06h00 was taken into account in fair usage enforcement.

If you were not incurring any usage during nightsurfer time, you get a bit less usage for a lot less money, if you were scheduling downloads for nightsurfer time, you get a lot more usage for a lot less money, if you were running at constant speeds 24 hours a day, you get about the same usage for much less.

It is unfortunate that this wasn't explained more explicitly at the time, but we hope to have some information on tracker at some point.

Why are we doing this? To move less valuable traffic (e.g. p2p) away from times when real-time traffic (e.g. streaming, video-conferencing) is more important, but (ideally) without making p2p unuseable during peak times (which is one of the other options, with its own problems).
Click to expand...
That is the case presently. Daytime torrents are throttled quite badly.

Now you want to make us use the midnight run.

I don't think this is going to be popular. Look also at the top users of other ISPs (yes, they list them). Compared to them we're download amateurs.
 
Taranis said:
That is the case presently. Daytime torrents are throttled quite badly.

Now you want to make us use the midnight run.

I don't think this is going to be popular. Look also at the top users of other ISPs (yes, they list them). Compared to them we're download amateurs.
Click to expand...
How much can you DL in 6 hours on a 40Mbps service (x30 for the month)?
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X