Prev: hostname again
Next: Using same TCP ports in diffferent zones
From: bugbear on 8 Apr 2008 07:29
I'm still trying to track down the cause of my nightmare.

To briefly recap, if I have "a few" Java threads,
and they shell out, I get woeful performance,
and massive idle time (as show by top).

I have found (RTFM!) a wondrous tool call plockstat,
and it appears that I have some kind of lock contention
issue.

However, the output of plockstat is very difficult
(for me) to understand.

I *think* I'm getting a summary for every process,
which since I'm shelling, means a lot of stuff.

For example, on a run where I have 8 threads each trying
to do 25 shell outs, I get 222 "Mutex block" headings
in my plockstat file.

Any help in interpreting this file and/or
analysing my contention issue would be gratefully
received.

BugBear
From: Helmut Kreft on 8 Apr 2008 18:41
["Followup-To:" header set to comp.unix.solaris.]
On Tue, 08 Apr 2008 12:29:34 +0100, bugbear wrote:
> I'm still trying to track down the cause of my nightmare.
>
> To briefly recap, if I have "a few" Java threads,
> and they shell out, I get woeful performance,
> and massive idle time (as show by top).
>

Have a look at this document:

http://docs.sun.com/app/docs/doc/816-5137/gen-12013

This may explain the trouble your're in.

Helmut

--
Almost everything in life is easier to get into than out of.
(Agnes' Law)
 | 
Pages: 1
Prev: hostname again
Next: Using same TCP ports in diffferent zones