dtrace discuss - Oct 2008 - Can you explain ''elapsed times'' from the ''procsystime'' script?

Hello,
Across on ''storage-discuss'' I''m trying to analyse a
problem involving
the Solaris iSci target, and thought that this would be a good
opportunity to try DTrace.

Having read some blog posts, I decided the easiest way
to start would be to use Brendan Gregg''s script here:

 http://www.brendangregg.com/DTrace/procsystime

This runs fine, and I have posted the results from
running it on a couple of my systems:

http://mail.opensolaris.org/pipermail/storage-discuss/2008-October/006458.html

http://mail.opensolaris.org/pipermail/storage-discuss/2008-October/006459.html

The ''Syscall Counts'' look ok, but I do not understand
the ''Elapsed Times'' statistics.

These claim to be in nano Seconds. Nano is 10^-9 or 0.000000001

So, using the figure from the first test in the second link above,
that makes the "Elapsed Times Total" of
2194329238161 correspond to 2194.329238161 seconds.

But I only had the script running for about 90 seconds,
before I did a Ctrl-C.
I don''t understand that!
Am I missing something? Can someone please explain?

Next Question:
I thought (in my naivety), that I would be able to look up further
information on the syscalls here:
http://docs.sun.com/app/docs/doc/819-2241/6n4huc7i6?a=expand
... but nothing there on ''kaio'', ''recv'',
''pollsys'' and ''lwp_park''.
Are these ''syscalls'' documented any where?

Many thanks, in anticipation of some advice.
I''m trying to learn some DTrace, but as you can see,
I''m currently at an early stage in that learning :-)
Best Regards
Nigel Smith
-- 
This message posted from opensolaris.org

Sorry, I dont know much and am on a mobile device... But any chance you are
doing this on a multi core/proc system ? On a multi processing system cpu time
can be more than wall time.
-- 
This message posted from opensolaris.org