uneducated attempt at statistical analysis
You can download the python script and excel files I used from here (the script outputs a csv file):
http://www.filehost.co.za/file/835-5588683304.html
I used a sample size of 100 people. I generated random numbers for daily bandwidth use according to these percentages:
Lowend: 25% 1-3GB per day
Average: 50% 3-10GB per day
Aboveavg: 20% 10-30GB per day
Abusers: 5% 30-40GB per day
In sample 1 the users stay in the same group, so continue with normal usage patterns when unshaped. In sample 2, the a few users start to use lower amounts of bandwidth as the month progresses...
Shaping starts at the end of day 1, with the top 50% of users being shaped. The top 50% are calculated on a 10 day rolling total (which starts at one day and builds up to ten…).
Sample 1:
In this sample day 1 usage was 960GB.
The rest of the days fluctuate between: 203GB and 482GB
Days 2 – 9 average 239GB
Days 23 – 30 average 253GB
Lowend: 2 shaped one day each; average was 59GB for the month.
Average: every user shaped 16.4 days per month; average was 87GB for the month
Aboveavg: every user shaped 24.65 days per month; average was 111GB for the month.
Abusers: every user shaped 27 days per month; average was 102GB for the month.
Next, I adjusted the numbers so that over the course of the month more people started using less. By the 21st day:
Lowend adjusted to 45%
Average adjusted 45%
Aboveavg adjusted to 8%
Abusers adjusted to 2%
Sample 2:
In this sample, day 1 used 941GB.
The rest fluctuate between 146GB and 426GB. There are only 2 days above 300GB (each time the abusers get unshaped…).
Days 2 – 9 average 211GB
Days 23 – 30 average 167GB
A definite downward trend…
The following is based on the numbers at the start of the month (so if someone average changed to a lowend user they still get counted as average).
Lowend: every user shaped 3.8 days in the month; avg use was 52GB
Average: every user shaped 15.4 days in the month; avg use was 66GB
Aboveavg: every user shaped 22.3 days in the month; avg use was 80GB
Abusers: every user shaped 26.8 days in the month; avg use was 99GB
Problem with this model is that the users that switched weren’t necessarily the ones that used the most bandwidth (My tiredness is greater than what I think the difference in results would be...).
Conclusion: if users don't use less bandwidth as the month goes on then fewer people get shaped, and the shaping 'limit' stays pretty high. However the more likely scenario is that people will adjust how much they use after experiencing how bad the shaping is, so more people are likely to be shaped as the threshold slowly drops over the course of the month.
Notes: the minimum usage was 1GB per day, shaped users had 0GB per day. Average users were between 3-10GB per day, it's probably more realistic to use 0-10GB per day. this would lower overall usage = more people getting shaped. There are no people that increase their usage throughout the month, so maybe they can balance each other out?
The rolling cap is calculated once per day; if IS is doing it more often then that would affect the lower end users more (downloaded 2GBs at 6am, turned computer off, was shaped vs 1GB at 6am, 1GB at 6pm same day, not shaped).
I tried the same models with the top 20% being shaped. It lot fewer people ended up being shaped, but I didn't account for the lower line speeds so I think it looked a little kinder for IS...
tldr: IS suck
