Path: dp-news.maxwell.syr.edu!spool.maxwell.syr.edu!drn.maxwell.syr.edu!news.maxwell.syr.edu!newsfeed.icl.net!newsfeed.fjserv.net!news.mailgate.org!newsfeed.stueberl.de!proxad.net!freenix!oleane.net!oleane!newsfeed.berkeley.edu!ucberkeley!sn-xit-03!sn-xit-11!sn-xit-08!sn-post-01!supernews.com!corp.supernews.com!not-for-mail From: Sybrand Bakker Newsgroups: comp.databases.oracle.server Subject: Re: Oracle question from a Sys. admin, re: Solaris performance Date: Fri, 04 Mar 2005 00:34:53 +0100 Organization: Posted via Supernews, http://www.supernews.com Message-ID: Reply-To: sybrandb@hccnet.verwijderdit.nl References: <1109883311.791239.242560@f14g2000cwb.googlegroups.com> <1109890738.771661.110420@g14g2000cwa.googlegroups.com> X-Newsreader: Forte Agent 1.93/32.576 English (American) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Complaints-To: abuse@supernews.com Lines: 51 Xref: dp-news.maxwell.syr.edu comp.databases.oracle.server:237419 On 3 Mar 2005 14:58:58 -0800, tonij67@hotmail.com wrote: >My questions about monitoring performance is from an Oracle standpoint, >not Unix. I am not a DBA so I don't know how these instances are >running, if they are crawling or screaming I have no idea. > >I know how to use unix perf. mon tools; sar, mpstat, vmstat, etc...and >there is no paging going on. Yet. With 50 instances we often have >high CPU utilization but that is not accompanied by I/O wait and/or >blocked processes, so so far things are ok. These are 4 proc. V880s >with 32 GB of ram using EMC Clarion striped LUNs. > >I just don't know how far we can push it, there seems to be no end in >site as sales is on a rampage to get everybody and their brother to use >this product so for all I know we could be doubling ot tripling the >amount of instances before we see anything noticable. > >For example, one system has 53 instances and averages 33% CPU >utilization. Disk activity is heavy but there is no I/O wait or >paging. Is it a problem now? Not as far as I am concerned, speaking as >a system admin. Compare that to an older E450 with 2gb of ram that >spends all day at 100% with the run que in double digits and high I/O >wait...so far it seems the V880s are up to the task but I have no idea >where the breaking point is, and our DBAs cannot tell me either. Get acquainted with Statspack. Take two snapshots on level 5 within 15 minutes apart. Run $ORACLE_HOME/rdbms/admin/spreport specifying the 2 last snapshot id's upload the file to Anjo Kolks site http://www.oraperf.com The report will exactly tell you what the average transaction doing: using CPU, waiting for I/O or whatever. As soon as it is more waiting than doing anything useful (I have a situation where 99.88 percent of the time of a transaction is spent in waiting for disk), you have a problem. You simply must be I/O bound, or your databases have no end-users at all. If Mr. Garry's comment is adequate (and they usually are), and you are using 8.0 default parameters (which were way too small), your system is I/O bound *because* of the SGA being too small. If you are going to crank up all of them (which is usually the only thing many admins and DBA's 'tuning' a system do), you will have paging and swapping soon enough. -- Sybrand Bakker, Senior Oracle DBA