[Rathji]

Not that I am debating the veracity of your claim, but what tests are you running to enable you to determine what Shaw's routers are doing to the packets for a TCP connection for a specific process running on your system?

The fact that it is happening only at upload only just amplifies what I am saying. You have 500kb/s of upload, divided among 40 TCP connections most of which are trying to pump a continuous stream of data. Lets assume there are only 4 active torrents, and your vent connection and nothing else using any bandwidth, this means there is most likely 25 (6X4 +1) TCP connections sharing your 500kb/s upload.

500/25 = 20kb/s per connection since TCP shares all bandwidth between connections equally over a period of time. If your torrents are trying and occasionally successful at using more than their 20kb/s, it will cause buffers to fill and you will lose packets for your Vent. If you have extreme, then you would double that number to 40kb/s.

Since you are losing packets on Vent, it needs to send more packets (to make up for lost ones) which can only make the problem worse. To top this all off when a TCP connection loses ANY packets, it cuts the speed at which it is sending those packets out in half. This could clearly cause the effect you are describing.

I am not saying that this is happening, but I think it is far more likely than Shaw using DPI in Calgary to throttle torrents. I am totally open to the possibility that I am wrong on this, if someone can provide proof.