On Thu, 12 Jul 2007, Matthew Gregan wrote:
At 2007-07-12T09:18:34+1200, John Carter wrote:
Ah, I always confuse the names of these two packages. I was referring to
procps.
That's the fun thing about FOSS, it advances on such a wide front,
that there are always new bits appearing someplace when you're not
looking...
The procps package is the one that delivers all the very well known
old utilities like top, ps, vmstat, kill, killall.
What I found interesting is it has grown sometime, somewhen, when I
wasn't paying attention, nifty new ones have appeared / I'm reminded of
old ones I had forgotten.
eg.
pgrep - much shorter/accurate than my usual ps aux | grep blah
pmap - print maps.
watch - repeated run and watch output of command.
slabtop - Watch slaballocator. Interesting that. I'm sort of use to
watching the "top" and "vmstat" output and can now
automagically spot Bad Things happening in that. I haven't
a clue what sort of Bad Things to expect / look for in
"slabtop".
pwdx - pwd for other process.
I'm a tad pragmatic on this front.... I want whatever stats I can get from
Linux that will allow me to compare the resource consumption of two
algorithms.
Algorithms? Read the implementation, it should be relatively easy to predict
how they will behave time and space wise if you understand them. Once
you've done that, maybe you want to look at real numbers from runtime
experiments to confirm your predictions.
Yip. That's where I'm at. I believe that algorithm two is much
better... if I understand the Ruby interpreter correctly.
If the "View from the Operating System" tells me Algorithm 2 isn't
much better...then I'm about to learn something new and interesting
that I didn't know before...
A quick list of useful tools for performance/inspection work is:
The valgrind based tools are absolutely luvly. Absolute treasures in
any programmers toolchest... Alas, somewhat less useful in the cases of ...
* Long running very CPU intensive programs. (Too blooming slow)
* Garbage collected interpreted programs like Ruby scripts.
Kernel, syscalls, I/O: systemtap
Hmm. That's an interesting one I had missed. I'm glad I asked...don't
know what to do with it yet....but it's clearly an extraordinarily evil
solution looking for a bunch of really mean problems.
blktrace
Hmm. Interesting...
/sys/kernel/debug does not appear to be a debug filesystem
Hmm. Google gooogle, googly google...Hmm. I'm going to have to learn
what this debugfs thing is...
Frysk is intended to handle a larger set of monitoring tasks, but it's still
in development and probably of limited use (in its current state) to people
unfamiliar with it.
Hmm. More to digest. Could be very handy around here that one...
Ah well, the time finding and tweaking stopper wasn't wasted. None of
these so far would have given me the answer I wanted... but they
certainly will be useful in future.
Thanks!
John Carter Phone : (64)(3) 358 6639
Tait Electronics Fax : (64)(3) 359 4632
PO Box 1645 Christchurch Email : [EMAIL PROTECTED]
New Zealand
