Systemtap Performance Impact

I am looking for some real information about the impact of systemtap on performance. I know that static nuclear probes are documented somewhere on their website (with statistics), but what about probe probes?

I saw this post on my mailing list ( http://sourceware.org/ml/systemtap/2009-q1/msg00518.html ), which indicates the possibility of a one-step in some cases. Can someone explain / point out some resources that list which methods are used for which probes?

What service data will be provided process().function()? What about static tags?

+5
source share
1 answer

Here are some guidelines for overhead on a ppc64 machine. That is all I could find. It appears that SystemTap was designed as a very low overhead tracer in accordance with this article .

It seems that the user side of utrace and uprobe SystemTap is not where most of the work is done. Therefore, it should not differ greatly from the specified criteria for checking the kernel. Here are some utrace performance tests .

Since the current version is only 0.8, I think that there was not much work to quantify the performance and overhead of SystemTap.

+5
source

All Articles