RE: missed Anjo's webcast..

[Cary Millsap]

Title: RE: missed Anjo's webcast..

I agree wholeheartedly with your “no such thing as an ‘overall system’” comment. That is key. For the record, my view on this general topic includes the following observations:   * If you have a really high database buffer cache hit ratio (>99%), then you almost certainly have inefficient SQL in your application.   * From the hundreds of situations we’ve seen in the past two years (we help people analyze at least one Hotsos Profiler run per workday), not one single site would have improved their performance noticeably by caching their entire database in memory. (It’s Amdahl’s Law: not one performance problem we’ve seen has been caused by too much time spent waiting for PIOs.)   * Nearly every slow application we’ve seen since 1999 either spends the preponderance of its elapsed time doing too many LIO calls (which manifests itself as an apparent “CPU problem” or a “latch contention problem”), or too many inter-tier database calls (which manifests itself as an apparent “network problem”).   * The solution for 99%+ of performance problems I have ever seen or heard of is to reduce LIO count (fix SQL), or database call count (fix application code). The key is to figure out *which* SQL, *which* application code, and *what* to go do about it.   Cary Millsap Hotsos Enterprises, Ltd. http://www.hotsos.com Upcoming events: - Hotsos Clinic, Oct 1–3 San Francisco, Oct 15–17 Dallas, Dec 9–11 Honolulu - 2003 Hotsos Symposium on Oracle® System Performance, Feb 9–12 Dallas - Next event: NCOAUG Training Day, Aug 16 Chicago -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] On Behalf Of Gogala, Mladen Sent: Thursday, August 08, 2002 3:31 PM To: Multiple recipients of list ORACLE-L Subject: RE: missed Anjo's webcast..   Well, I guess that I disagree. Buffer hit radio does matter as one of the performance indicators, but certainly not the only one. Your and Mr. Milsap thesis is that LIO also is very expensive and its cost is far from being negligible, so having gazillion of LIOs instead of 100 times smaller number of PIOs will not make our system run faster. BHR alone cannot be used to judge to overall health of the system, but thebn again, there is no such thing as the "overall health of the system". It's the users of the system who will say whether the performance is satisfactory or not, and I'm usually tuning an application, not an imaginary "overall system". Low cache hit ratio usually tells me that I do have a hog who is using lots of PIOs. By my experience, it usually is a very good indicator that something is wrong, at least on an OLTP  system. So, after all, I do find BHR a useful indicator, but by no means the only one or the most important one. Event 10046, SQL_TRACE (level 1 of 10046), explain plan and v$session_event still are the tools I need most, but I still do need BHR  as an indicator. Mladen Gogala Oracle DBA Phone: (203) 459-6855 Email:  [EMAIL PROTECTED] -----Original Message----- From: Anjo Kolk [mailto:[EMAIL PROTECTED]] Sent: Thursday, August 08, 2002 1:05 PM To: Multiple recipients of list ORACLE-L Subject: Re: missed Anjo's webcast.. Moi wrong ;-) Jeeh, human after all   To summarize the webcast: db-block-buffers do mattter. Too many LIO do matter. Too many PIO do matter. But Buffer Cache Hit ratio doesn't matter ....... End user satisfaction does matter.   I am always willing to clarify any points that I made, you just have to ask me l ....   Anjo.     ----- Original Message ----- From: [EMAIL PROTECTED] To: Multiple recipients of list ORACLE-L Sent: Thursday, August 08, 2002 5:43 PM Subject: RE: missed Anjo's webcast..   Guys, I had this dream that I missed the webcast - which I did.  However, someone said it wasn't very interesting but the conversation of the people (gurus) left over was very interesting as there was good solid evidence that he was incorrect and db_block_buffers do matter.  Kind of inline with the discussion about redos yesterday and my indexing/partition issues - hmmm. -----Original Message----- From: Mladen Gogala [mailto:[EMAIL PROTECTED]] Sent: Thursday, August 08, 2002 1:38 AM To: Multiple recipients of list ORACLE-L Subject: Re: missed Anjo's webcast..   Www.precise.com, go to Events->webcasts... On 2002.08.08 00:53 Madhusudhanan Sampath wrote: > Are transcript documents available anywhere? > > Regards > Madhusudhanan S > > > > _________________________________________________________________ > MSN Photos is the easiest way to share and print your photos: > http://photos.msn.com/support/worldwide.aspx > > -- > Please see the official ORACLE-L FAQ: http://www.orafaq.com > -- > Author: Madhusudhanan Sampath >   INET: [EMAIL PROTECTED] > > Fat City Network Services    -- (858) 538-5051  FAX: (858) 538-5051 > San Diego, California        -- Public Internet access / Mailing Lists > -------------------------------------------------------------------- > To REMOVE yourself from this mailing list, send an E-Mail message > to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in > the message BODY, include a line containing: UNSUB ORACLE-L > (or the name of mailing list you want to be removed from).  You may > also send the HELP command for other information (like subscribing). > -- Mladen Gogala -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Mladen Gogala   INET: [EMAIL PROTECTED] Fat City Network Services    -- (858) 538-5051  FAX: (858) 538-5051 San Diego, California        -- Public Internet access / Mailing Lists -------------------------------------------------------------------- To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in the message BODY, include a line containing: UNSUB ORACLE-L (or the name of mailing list you want to be removed from).  You may also send the HELP command for other information (like subscribing).