> > Do you think is there a need to make the
> > gathering of Linux fine-tuning tips an easier
> > task for the novice end-user?
>
> Yes. Step one: Add performance metrics collection to the kernel. I've
> been working on this for a few weeks. I've got a good start on disk stats.
Did you look at Stephen Tweedie's sard work (from
ftp.uk.linux.org:/pub/linux/sct/fs/sard*) ?
> There was some talk about system call accounting last week. The cleanest
> place to add this is in assembly code which I'm not qualified to modify.
> (I don't find the number of system calls a particularly interesting number,
> but it should be there for correlation purposes.)
I think a far more interesting metric ATM is locking overhead. I have
a patch (based on original code from Andrea Arcangeli) that adds knowledge
about the locking segment to the time accounting code and outputs it in
/proc/stat. I also have some X11 hacks to display the locking overhead
graphically. Profiling spinlocks could be a very useful addition,
unfortunately it needs source changes to every spinlock setup.
Accounting system calls is not too useful I think, more interesting
would be to break the kernel time into "IO time", "bottom half time" (this
is currently accounted to some random process) etc. We don't want to
add too much cruft to the timer interrupt handler though.
-Andi
-- This is like TV. I don't like TV.- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/