NonprayingMantis:sorceror: here are some real world results (done just now from my UFB connection):
this made 16 TCP connections to reach that speed - clearly latency isn't an insurmountable issue anymore for speedtest.net
assume you are on a 200Mbps plan?
based on TimA's chart above, 16 TCP connections seems too low to get that result.
Ping of 172ms gives less than 5Mbps per connection (based on that chart), so 16 connections would, presumably, max out at less than 80Mbps, no?
(also, how do you know how many connections the speedtest.net test has done, I can't see it anywhere on speedtest.net? - the speedtest website claims to use up to 4 HTTP threads https://support.speedtest.net/entries/20862782-How-does-the-test-itself-work-How-is-the-result-calculated- although that 'article' is a couple of years old so maybe they use more now IDK
myfullflavour: This is on my 100/20 connection at home.
I get the best latency to Palo Alto / San Jose speedtest servers as our US upstream peers at layer3 in that area.
Los Angeles is obviously further down the coast.
sorceror: i get better latency than that to Palo Alto (in the mid 120s) but I can't get over 20M to either of the speedtest servers there - just goes to show (again) that latency is not very useful when it comes to estimating your max throughput from speedtest.net
Any comments made are personal opinion and do not reflect directly on the position my current or past employers may have.