one plus one
Shared publicly•View activity
- bob bowlesModeratorIt's not a bad thing to test with audio that is more active than real world. That might lead to the real world working better than your tests! (And how often does that happen on Internet scale :-)May 8, 2014
- true. As I wrote, it's only a quick estimation test. Nobody usually runs a conference with hundreds of participants. For a more thorough testing, one needs multiple smaller rooms (which is actually easy, "xx" is the room name in that article)May 8, 2014
- bob bowlesModerator+1We (ie: smart colleagues of mine) have automated scale tests using SIPP that does just that ('rooms' of all sizes across many cores and threads). The 'canary in the coal mine' is a single room playing classical music with one human listener. Even though all the other streams are fully instrumented and lots of data collection occurs, there is nothing like the human ear to pickup that first imperfection as the hardware crests its performance limit.May 8, 2014
- thanks, I updated the article. Also see my emailMay 8, 2014
Add a comment...