Path: news.f.de.plusline.net!news-fra1.dfn.de!news.tele.dk!feed118.news.tele.dk!postnews.google.com!h2g2000hsg.googlegroups.com!not-for-mail From: hpuxrac Newsgroups: comp.databases.oracle.server Subject: Re: Kernel parameter recommendations for HP-UX 11.11 in regard to SGA size Date: Wed, 06 Jun 2007 06:58:13 -0700 Organization: http://groups.google.com Lines: 63 Message-ID: <1181138293.904649.252940@h2g2000hsg.googlegroups.com> References: <4665340f$0$22417$afc38c87@news.optusnet.com.au> <1181089675.265857.272630@k79g2000hse.googlegroups.com> <1181091566.954412.246060@i13g2000prf.googlegroups.com> NNTP-Posting-Host: 70.228.67.48 Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Trace: posting.google.com 1181138294 28805 127.0.0.1 (6 Jun 2007 13:58:14 GMT) X-Complaints-To: groups-abuse@google.com NNTP-Posting-Date: Wed, 6 Jun 2007 13:58:14 +0000 (UTC) In-Reply-To: <1181091566.954412.246060@i13g2000prf.googlegroups.com> User-Agent: G2/1.0 X-HTTP-UserAgent: Mozilla/4.0 (compatible; MSIE 7.0; Windows NT 5.1; .NET CLR 2.0.50727),gzip(gfe),gzip(gfe) Complaints-To: groups-abuse@google.com Injection-Info: h2g2000hsg.googlegroups.com; posting-host=70.228.67.48; posting-account=Dz_3bQ0AAAC5LZNB2NPEJz1sl_a8qcDY Xref: news.f.de.plusline.net comp.databases.oracle.server:199014 On Jun 5, 8:59 pm, joel garry wrote: > > Since it probably looks like I'm giving opposite advice to this, I > should say, any time hpuxrac and I appear to disagree about hp-ux, > listen to him :-) There's no shortage of people giving bad advice here on cdos either intermittently or more frequently. Take each piece of advice on a case by case basis, do the research, and build test cases to support arguments one way or another is I think advice that is more appropriate Joel. > In my customer's case, it is a fast set of processors with a not-so- > fast RAID-5, and happens to work well with defaults (at least, no > complaints - I have seen the hardware get behind on massive loads). I > would rather do it the "correct" way, but there you go. It doesn't > seem to be wasting memory, I would like to know how one might tell > without glance (an optional utility). I understand about double > buffering being wasteful, but I also understand about tuning that > which is not a bottleneck being useless. I tend not to think of either glance or sam being "optional" on an hpux system. With glance you can click on the Memory Report then look at the bottom of the screen for Buf Cache to see what is currently allocated. My experience is that hpux tends to strongly bias itself up to allocating huge amounts of memory ranging up to dbc_max_pct. Anecdotally some people seem to think that having more than a gig of memory allocated doesn't hurt their systems but it sure doesn't make much sense to me. If you have a system with 8 gig of memory I would think seriously about bumping down dbc_max_pct from the default. At least check out periodically what is being used. If you don't have glance handy I think there are some other shareware/freeware reporting tools ( is it freemem ? ) that might be useful. Yes probably there's a command line to get that information also. > > Swoop, Metalink Note:401323.1 helps clarify the vfxs options. > > > > Probably 256 meg or less ( 128 maybe ) if is adequate so figure out 4 > > gig times what number gives you that and set both dbc_max_pct and > > dbc_min_pct to the same value. Looks like some people on the hp forum like to set the values much higher. Probably opinions vary widely. To the OP, don't take my 256 shot in the dark as gospel for your system. > > > The other ones I would have to double check at work tomorrow but I > > think they both should be zero. Yes they are both zero on our systems which apparently is designed to let hpux go between the 2 values dynamically for dbc_min_pct and dbc_max_pct. My opinion though is to consider setting those values fairly close if not identical.