Xref: utzoo comp.sys.hp:4799 comp.unix.wizards:21394 comp.databases:5534 comp.unix.questions:21327
Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!iuvax!cica!tut.cis.ohio-state.edu!cs.utexas.edu!uunet!munnari.oz.au!kaukau.comp.vuw.ac.nz!comp.vuw.ac.nz!dsiramd!actrix!paul
From: paul@actrix.co.nz (Paul Gillingwater)
Newsgroups: comp.sys.hp,comp.unix.wizards,comp.databases,comp.unix.questions
Subject: What could be causing huge number of semops?
Message-ID: <1990Apr10.211715.27253@actrix.co.nz>
Date: 10 Apr 90 21:17:15 GMT
Reply-To: paul@actrix.co.nz (Paul Gillingwater)
Organization: Actrix Public Access UNIX, Wellington, New Zealand
Lines: 18

Hi,
I'm working on an HP-UX 3.1 system with, inter alia, Informix's
C-isam software.  Informix appears to make fairly heavy use of
shared memory and semaphores.

The box is under heavy loading (over 130 users -- it's an HP9000/855)
and we've done substantial kernel tuning to eliminate bottlenecks,
but this last one has got us baffled.

At certain times, the semaphore activity gets VERY high.  Using sar -m
shows sustained averages of over 500 semops per second.

Has anyone seen this behaviour?   What effects might it have?  How
can we try to reduce the impact of this semaphore usage?

Please e-mail, I'll summarise if anything useful arises.
-- 
Paul Gillingwater, paul@actrix.co.nz