[PATCH RFC] Summary of syscall latency
Mark.Hills at framestore.com
Tue May 27 16:44:35 UTC 2014
On Tue, 27 May 2014, Dmitry V. Levin wrote:
> On Tue, May 20, 2014 at 10:26:11AM +0100, Mark Hills wrote:
> > Time spent in system time is not useful where a syscall depends on some
> > non-CPU resource, eg. typically open() or stat() to a network drive.
> > This patch adds a -w flag to produce a summary of the time difference
> > between beginning and end of the system call (ie. latency)
> > This functionality has been useful to profile slow processes that are not
> > CPU-bound.
> > An older commit 8050cdc mentions this:
> > > It might be useful to have a mode where we show wall clock time spent in
> > > syscalls, not CPU time. It might also be more accurate.
> > I'm not sure how reasonably useful the crude profile of CPU "system time"
> > is. Is it better to replace the existing functionality with
> > wallclock/latency time, or does this justify the introduction of a new
> > flag, as below?
> I think a new behavior justifies a new flag. Another option is to treat
> the same -c/-C option given twice as a request for this new behavior.
In which case the -w flag seems reasonable; given the two existing flags I
think the variations would be confusing for -c/-C given twice.
I'll prepare the final patch and send it. I assume I should add
Signed-off-by and a NEWS entry.
More information about the Strace-devel