@wycats It's the contrast with "everyday" that breaks down. Many 95ers get that performance everyday. I'd go with Most/some/few or the like.
-
-
-
@dhh I'm thinking "Normal Requests" / "Problem Requests" (95% happens many times per day, just as 95% rain happens many times per year) -
@wycats Normal requests taking 500ms are a "problem". 95s taking 200ms probably not. -
@dhh Right, but 95s taking 2s probably is, even if median is 200ms, and this is common. -
@wycats There's your answer. A problem request is one that takes 2s, whether it happens at the median or the 95th. -
@dhh not totally obvious. Tolerance is different at different occurrences. -
@wycats By that token labeling 95ths "problem" isn't apt either, though. And 2s at the median would be 'catastrophe'. -
@dhh yeah the hard thing is labelling relative occurrences so 95p sounds common but not typical. - 6 more replies
New conversation -
-
-
@wycats only if you assume a normal distribution -
@mattlarraz@wycats Regardless of distribution, 50th %ile is everyday perf. Half see better, half see worse. -
@mattlarraz@wycats The only thing that would depend on normal distribution is if you're comparing median with average. - End of conversation
New conversation -
-
-
@wycats If you can identity correlations, the labeling problem goes away. Not "outliers took 10s", but "users on EDGE networks took 10s" etc -
@wycats the more abstract category is hard to label precisely because it's not very meaningful by itself. Needs next-level analysis.
End of conversation
New conversation -
-
-
@wycats I like to think about it in absolute numbers. Can I live with having 1000 people leaving due to poor performance?Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@wycats Median is not everyday. Half your users get worse than median.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@wycats 95th percentile are still people. Waiting.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.