Hi, On 11 May 2017 at 08:27, Guruganesh Hegde <guruhegde4u@xxxxxxxxx> wrote: > > When fio is run with blktrace, > IOPS drops by 8%, average clat increased by 9% and max clat is5 times > more compared to the test run without running blktrace. > > Is this expected overhead when blktrace is run? > Wondering why max latency is 5 time more!! This isn't an easy question to answer as there are too many variables: How much CPU is left when fio is running by itself? How much is left when you are running it and blktrace together? What sort of IOPs are you seeing without blktrace? Are you saving your blktrace log to a completely different disk on a completely different controller to the ones being used by fio or otherwise storing the entirety of the trace entirely in RAM? If you generate competing I/O towards the same disk(s) you'll hurt your speeds. Also this isn't so much an fio question as a blktrace one so it might also be useful to ask on the blktrace mailing list (see the bottom of http://git.kernel.dk/?p=blktrace.git;a=blob;f=README;h=f52f48d8d621f916d7b940bbe5ed2b9b92f4c836;hb=43fc870ce04e963def45dfc0d1ed4ea21ef10d4b#l172 ). -- Sitsofe | http://sucs.org/~sits/ -- To unsubscribe from this list: send the line "unsubscribe fio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
