This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
Re: Questions on Systemtap performance
- From: "Frank Ch. Eigler" <fche at redhat dot com>
- To: shan zeng <zengshan2008 at gmail dot com>
- Cc: systemtap at sources dot redhat dot com
- Date: Fri, 1 Jan 2010 11:47:47 -0500
- Subject: Re: Questions on Systemtap performance
- References: <14b21e1d1001010721r22398527vb8af6fe7ca928110@mail.gmail.com>
- Reply-to: systemtap at sources dot redhat dot com
Hi -
> When using SystemTap to probe in the kernel functions to get the
> executing time of the function, there must be performance cost when probing,
> how can I quantify the performance and overhead of SystemTap? Is there any
> tool which can do the job?
The basic overheads arise from the kernel's implementation of
kprobes/kretprobes, which on modern hardware tends to be in the
microsecond range per hit. (In future kernels, this should be reduced
to about 25% with "djprobes".)
Systemtap can measure its own generated code's overhead (stap -t), but
does not measure the kernel's inherent overheads yet.
- FChE