Don"t take things like this too seriously, capacity testing, especially with multiple concurrent flows is a bit tricky and especially the intermediate values are prone to show gunk...
Yes. it would be nice if that would be different, but alas it is not...
1 Like
Fair, it took me a while to notice the pattern (which can be more or less pronounced), and talking to the wavefront guys made me appreciate how much they need to work against the browsers for these latency measurements...
1 Like
To illustrate my point:
waveform bufferbloat test on macos firefox:
and on macos safari:
Note the distinct latency bands in the 250-310 range (and a bit wider for idle)? These have nothing to do with the network and a lot with the browser. Sure Firefox is not squeaky clean either, but 2 outliers while likely also driven by the browser do not skew the averages while for safari it does, except the idle measurement was also skewed...
The take home message is: modern browsers are marvellous tools, that like a Swiss army knife can do a lot of things, but measurement environments for high precision data acquisition they are not...
1 Like