I have decided to split this issue out from the "MWEB Uncapped Subscribers Feedback" on account of it becoming lost amongst a host of other unrelated issues. The severity of the issue, coupled with MWEB's apparent refusal to acknowledge that there is a problem at all has forced me to try and bring the issue to the forefront in an attempt to put pressure on MWEB to rectify it.
What is the issue?
Currently, Durban and Johannesburg users of MWEB's Uncapped Shaped product are being throttled on a per connection basis (regardless of protocol) during the day. This is on a sliding scale, but at its worst is around 150kbps, or 18KB/s. This phenomenon is most clearly illustrated in the graph below:
http://www.slackers.co.za/uploads/20110525/daily.gif
This graph shows the progression of a very large download via HTTP (port 80) from a web server in Europe.
You may ask why this is a problem, given that we are talking about a shaped product, and we can expect shaping during peak hours of the day... You'd be justified, and if every person that paid R539 for the service got the same performance, there would be no room for complaint. However, it is a problem because users in Cape Town are most definitely NOT subject to the same shaping / throttling. At any given point in the day, User A in Cape Town can expect up to 10x the performance when compared to User B in JHB/Durban!
What is MWEB's response?
From the responses in the "MWEB Uncapped Subscribers Feedback" thread (which incidentally has grown by almost 100 pages since this problem began) with regard to users experiencing the same problems, the general concensus from Will@MWEB and MWEB Guy is that there are numerous external factors that need to be considered, and that we can be assured that they do endeavour to keep the experience the same in all regions.
I, for one, could accept this explanation if we were talking about a fluctuation of 10-20%, but when we are talking about such a MASSIVE difference in performance, questions need to be asked about what is a fair expectation!
I have personally spent many hours testing out this theory by comparing my results with those of some 10 other MWEB Uncapped users (some high-usage, others low-usage), both in JHB, Durban and Cape Town which is why I am so confident that the issue does not lie with my line or my specific account. Unfortunately, at this point in time I have basically been stone-walled by MWEB support who refuse to entertain any further support queries from myself on the matter given my high usage. Apparently, because I have downloaded such large volumes of data in the past, I am not in a position to complain about my throughput. As such, there really is not much more I personally can do, other than try to bring the issue to the attention of others and hope that someone else takes the baton and runs with it. Hopefully a tech-savvy *new* MWEB client in Durban will pick up on this and query it with MWEB. That way, there can be no question of usage.
As I did in the original thread, I would like to challenge ALL 4MB MWEB Uncapped users in the Durban area to try and download a file of 20MB or more at an average rate of more than 75KB/sec using a single connection (so no download managers) during the week-day. Time your download and work out what the average speed was at the end (don't look at the realtime speed). Then, give the same URL to a user in Cape Town and compare the speed. If you don't know a user in Cape Town, hopefully someone reading this can put their hand up to test for us. I suspect you will find that the user in Cape Town will get the same file at least two to three times, and up to TEN times as fast. Do note however, that many international downloads are hosted on CDNs with local mirrors, or are cached locally. Therefore, to get a true reflection of your download speed, try to use a unique URL (ie. your own webspace in US or EU, or Dropbox etc), or perhaps choose an international mirror for a file listed here.
When I can find the time, and if I am still an MWEB customer at that point, I would like to write a small app that will allow MWEB users to download a unique file on a web server and compare their throughput to that of others, as well as collate that information. I'm confident that this would highlight the huge discrepancy in performance between JHB/DBN and Cape Town.
UPDATE:
User Pr⊕phet has found a nice speedtest site that clearly highlights the issue above. I have created a public Google Spreadsheet to keep track of the results. It can be viewed here.
What is the issue?
Currently, Durban and Johannesburg users of MWEB's Uncapped Shaped product are being throttled on a per connection basis (regardless of protocol) during the day. This is on a sliding scale, but at its worst is around 150kbps, or 18KB/s. This phenomenon is most clearly illustrated in the graph below:
http://www.slackers.co.za/uploads/20110525/daily.gif
This graph shows the progression of a very large download via HTTP (port 80) from a web server in Europe.
You may ask why this is a problem, given that we are talking about a shaped product, and we can expect shaping during peak hours of the day... You'd be justified, and if every person that paid R539 for the service got the same performance, there would be no room for complaint. However, it is a problem because users in Cape Town are most definitely NOT subject to the same shaping / throttling. At any given point in the day, User A in Cape Town can expect up to 10x the performance when compared to User B in JHB/Durban!
What is MWEB's response?
From the responses in the "MWEB Uncapped Subscribers Feedback" thread (which incidentally has grown by almost 100 pages since this problem began) with regard to users experiencing the same problems, the general concensus from Will@MWEB and MWEB Guy is that there are numerous external factors that need to be considered, and that we can be assured that they do endeavour to keep the experience the same in all regions.
I, for one, could accept this explanation if we were talking about a fluctuation of 10-20%, but when we are talking about such a MASSIVE difference in performance, questions need to be asked about what is a fair expectation!
I have personally spent many hours testing out this theory by comparing my results with those of some 10 other MWEB Uncapped users (some high-usage, others low-usage), both in JHB, Durban and Cape Town which is why I am so confident that the issue does not lie with my line or my specific account. Unfortunately, at this point in time I have basically been stone-walled by MWEB support who refuse to entertain any further support queries from myself on the matter given my high usage. Apparently, because I have downloaded such large volumes of data in the past, I am not in a position to complain about my throughput. As such, there really is not much more I personally can do, other than try to bring the issue to the attention of others and hope that someone else takes the baton and runs with it. Hopefully a tech-savvy *new* MWEB client in Durban will pick up on this and query it with MWEB. That way, there can be no question of usage.
As I did in the original thread, I would like to challenge ALL 4MB MWEB Uncapped users in the Durban area to try and download a file of 20MB or more at an average rate of more than 75KB/sec using a single connection (so no download managers) during the week-day. Time your download and work out what the average speed was at the end (don't look at the realtime speed). Then, give the same URL to a user in Cape Town and compare the speed. If you don't know a user in Cape Town, hopefully someone reading this can put their hand up to test for us. I suspect you will find that the user in Cape Town will get the same file at least two to three times, and up to TEN times as fast. Do note however, that many international downloads are hosted on CDNs with local mirrors, or are cached locally. Therefore, to get a true reflection of your download speed, try to use a unique URL (ie. your own webspace in US or EU, or Dropbox etc), or perhaps choose an international mirror for a file listed here.
When I can find the time, and if I am still an MWEB customer at that point, I would like to write a small app that will allow MWEB users to download a unique file on a web server and compare their throughput to that of others, as well as collate that information. I'm confident that this would highlight the huge discrepancy in performance between JHB/DBN and Cape Town.
UPDATE:
User Pr⊕phet has found a nice speedtest site that clearly highlights the issue above. I have created a public Google Spreadsheet to keep track of the results. It can be viewed here.
Last edited:
video at 720p with no buffering.