09:30 <lightydo> I have been looking for information on the effect/overhead of systemtap on a production system and found this: https://sourceware.org/ml/systemtap/2006-q3/msg00146.html

09:30 <lightydo> is there anything more updated ?maybe a guide on running similar tests on a system to get an idea on the performance hit introduced by my systemtap scripts?

09:44 <mjw> yeah, there probably have been some changes in the last 12 years. I know jafeer did some benchmarks recently, but I cannot find them. Maybe fche knows if those have been published.

12:58 <fche> lightydo, perchance we should include some formal self-benchmarking scripts in the package

12:58 <fche> usually an effective way of measuring impact is to multiply rate-of-events by cost-of-events :)

12:59 <fche> the former factor is under control of the stap script and the workload; the latter is under control of the kernel & stap runtime

12:59 <fche> so ... quite a few degrees of freedom. but a benchmark script can estimate at least a few of the more fixed values

14:49 <lightydo> I am building some software that probes and analyzes various syscalls and I want to know if I run it on say a webserver, how my software impacts the performance of that server...

15:06 <fche> lightydo, understood, given the large number of variables under only your control, often the best way to answer that question is to measure for yourself