Earlier quoted context omitted.
Try a single threaded download from Hetzner Finland versus eu-north-1 to a remote (i.e. Australia) destination and you'll see premium bandwidth is very real. Google Cloud Storage significantly more so than AWS. Sure you can just ram more connections through the lossy links from budget providers or use obscure protocols, but there's a real difference. Whether it's fairly priced, I suspect not.
I just tested it and TCP gets the maximum expected value given the bandwidth delay product from a server in Falkenstein to my home in Australia, from 124 megabits on macOS to 940 megabits on Linux. Can you share your tuning parameters on each host? If you aren't doing exactly the same thing on AWS as you are on Hetzner you will see different results. Bypassing the TCP issue I can see nothing indicating low network qu…
Re: A $1k AWS mistake
#281Now try when you don't control the host to tune on.