Re:How Not to Measure Latency [pdf]
by hyperpape on Sunday November 01, 2015 @02:55PM
(#10487880)
Thanks for the paper!
On classifying testing tools as open/closed, I've been able to use JMeter to simulate open requests against very heavy endpoints [1] by not having individual threads loop, increasing the number of threads and using the ramp-up feature.
I suspect that this wouldn't work for testing something that can handle large amounts of traffic, but there are cases where you can fit a squarish peg into a somewhat round hole.
[1] 100s of ms or seconds of time for endpoints that do a lot of work (both CPU and IO) per request).
On classifying testing tools as open/closed, I've been able to use JMeter to simulate open requests against very heavy endpoints [1] by not having individual threads loop, increasing the number of threads and using the ramp-up feature.
I suspect that this wouldn't work for testing something that can handle large amounts of traffic, but there are cases where you can fit a squarish peg into a somewhat round hole.
[1] 100s of ms or seconds of time for endpoints that do a lot of work (both CPU and IO) per request).