Hacker News new | ask | show | jobs
by datavirtue 1737 days ago
Can someone bake this down to a sentence? I think I understand what they are saying since I have been faced with using these metrics and in having been tempted to use average response times I recognized that average is not a good baseline since it moves in relation to the outliers (which are usually the problem requests and there can be many or few).

How do you calculate these percentiles and use them as a trigger for alerts?