TPG: a scourge to society
/According to their website, “TPG is a leading Australian ISP Telco providing cost effective and reliable NBN, ADSL2+, Fibre, DSL Internet Broadband with Phone and BYO Mobile plans”. In my experience, TPG is the bottom rung of the Australian internet ladder - which is no mean feat - and unable to deliver an acceptable quality of internet service. Moreover, their ability to gag their own support staff and find seemingly the least competent technicians and engineers is simply astounding. Unfortunately, what adds the salt to this festering wound is the state internet infrastructure in Australia, which has been purposely crippled by commercial interests to the detriment of all. The result is that you have no choice but to accept your lot. During my fight for a better internet connection, I made a few scripts which might be useful to others and also produced some neat plots, so if that is your thing please, take a look around.
A national broadband network to connect all Australian homes to an all-fibre network was a great idea and necessary step forward. So when it began to crumble it was disconcerting, and over the following years has become an issue I can barely talk about; the anti-competitive venom and blatant corruption that resulted in the network being neutered is sickening. I very real consequence of the policy is that despite living some five kilometres from the heart of Melbourne, I am on an Asymmetric Digital Subscriber Line (ADSL) 2+ connection which is quite literally held together with tape and glue: the connection is dependent on the weather conditions along with the whims of the gods. Making matters worse, our internet provider TPG is the worst in the business. Their model is to provide bare-bone systems with no support, which is great when a) you are competent with networking and b) your infrastructure has no major faults. You can work on the former but the latter is harder. In any case, the quality of our internet took a sharp turn for the worse in March of 2019. Consulting with TPG support was akin to getting cosy with a grater, but the consistent theme over many interactions is that they simply did not believe that the quality of my connection was not good, simply stating that it looked “above average”. It was about this time I thought I would simply write some code to monitor the quality of my connection and then go from there.
The code
A repository of the code can be found here, but the basic premise is to try and connect to Google and if there is a response, I deem this “online”. If there is no response, I will say that I am “offline”. By design, it is not sophisticated, just something simple what will be run on a minute-by-minute basis so statistics can be compiled.
The results
When I first got the code and analysis working, I was pretty amazed to see the output. Shown below is the first plot I produced, a 24 hour period centred on 0200 on the 27th of March, 2019.
To better explain the plot, the x-axis shows the time (one point every 60 seconds) and there is a binary output of either online or offline. The blue line shows the raw data and the purple shows the same data, but only when consecutive values are the same, simply to remove spurious data points, which I would argue is being very generous to TPG, suggesting that isolated points must be spurious rather than real.
It was not long after providing TPG with this plot (along with others) they sent a service technician to our property - despite maintaining that there was no issue with our service - who said little more than “yeah, seems like something is fucked”. This was not a particularly satisfying result, but it was in the process of troubleshooting this that I noticed a correlation with a capped upload speed and the stability of our internet connection. Our upload speed had always been atrocious (usually around 0.1 Mbps) and by throttling our upload to 10 kilobytes/second we gained a more usable internet connection; however good luck trying to back anything up!
Accepting this as our Modus operandi was annoying but it did the job - mostly. There were periods of frustration, to the point of buying a 4G dongle and debating setting up a direct-wireless link to a friend with fibre to the basement roughly one kilometre away. But it was not until we moved out of that property and happily severed our connection with TPG that I bothered to look into the data produced from the code. I had set it up to prove to TPG there were issues and just left it running. We moved out of our abode on January 9th, 2020 and so shown below is our entire history of internet connectivity with TPG.
TPG connectivity: the complete story
You can see that there were periods of good and bad, notably good in August and October, which is when we were on leave and there was no network traffic. The bad periods came and went, but nothing was worse than late September - which coincidentally was when I cracked and purchased the 4G dongle. Quantifying the percentage of time connected by month
2019: March 0.940
April 0.945
May 0.938
June 0.962
July 0.965
August 0.983
September 0.895
October 0.982
November 0.983
December 0.983
2020: January 0.958
one can see that 96% connectivity is about the average, but September dipped below 90%. Looking more closely at this time period, the percentage connectivity by week are listed below
2019-09-15 0.977
2019-09-22 0.884
2019-09-29 0.770
2019-10-06 0.965
which shows that things at the end of the month were pretty dire. Below, I have plotted the connection log as before for the period of September 20th through 30th to get an idea of how things looked.
In contrast to some of the previous plots, there are extended periods of being offline and only spikes up to offline, especially in the period of September 25th and 26th. So if you are an executive trying to flog more 4G dongles and wondering how to improve sales, make home internet connectivity look like this. It will break people pretty quickly.
More than just looking an binary online and offline, it is also possible to look at the percentage of time connected when the data is binned into time divisions. In these cases, I have chosen plots with bins of one day and one hour respectively, which are shown below.
As one would expect, the data is averaged for with longer bin widths, but even still there are periods which drop to 50% connectivity per day, whereas when looked at on a per hour basis, there are many times when there connectivity grinds along at 0%. How good is the internet in Australia?
A post script
Moving not only provided a good opportunity to change internet providers, but also gave us a chance to change infrastructure: we moved into a house which was part of the original NBN rollout, meaning it has fibre to the home. Moreover, rather than joining with TPG, we joined with Aussie Broadband, which at the time had the highest customer satisfaction rating of any provider in the country. In addition to ripper speeds, below are some plots that make me pretty happy.