Oracle-L: Re: mbrc (was: buffer cache

From: Boris Dali <boris_dali_at_yahoo.ca>
Date: Tue, 13 May 2003 14:06:46 -0800
Message-ID: <F001.005982AA.20030513140646@fatcity.com>

Wolfgang,

Wolfgang Breitling <breitliw_at_centrexcc.com> wrote: > I suppose with mbrc you actually mean dfmrc
> (db_file_multiblock_read_count).

Yes, thank you (fat fingers, long to type ...)

> In 9i there is a
> system_statistics value
> MBRC which is used by the CBO to cost full scans in
> place of dfmrc,
> provided you have collected system statistics. You
> can then set dfmrc to
> the highest value possible for your OS without
> incurring more full scans.

Thanks. I'll look into this

>
> As for your problem, is it possible that you just
> hit a busy spot during
> your dfmrc=128 setting, or are those results
> repeatable? In the latter case
> it may be an efficiency deficiency on the part of
> the OS: maybe it is
> taking much longer to read 128 blocks than it does
> reading 16 blocks. What
> are the results for interim values - dfmrc=32, 64,
> 96 ?

At the time I ran the tests - yes, I could easily reproduce it (meaning it was repeatable) I'll do some additional tests in off-hours and also with different dfmrc values as you suggested. I will also be able to test it on a different machine and different OS

My block is 8K (no multi block tablespaces/buffer pools) and I am on ext3. Can it be something to do with how ext3 is configured? What's filesystem block size on ext3?
I kinda miss Solaris's tunefs, fstyp, kstat (so I don't know where to look for maxcontig, filesystem buffer cache - assuming they exist and tunable on ext3). I am also not sure if ext3 has something similar to Solaris UFS write throttling. Instead however I've got elvtune to (presumably) trade throughput for latency (or visa versa). Has anybody tried messing with all this? Any study on how this can be related to dfmrc? Or is it suppose to be the higher dfmrc the better (for multiblock access patterns and assuming CBO doesn't get affected)?

Thanks again, Wolfgang.

Boris Dali.

Post your free ad now! http://personals.yahoo.ca

-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.net
-- 
Author: Boris Dali
  INET: boris_dali_at_yahoo.ca

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services
---------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (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).

Received on Tue May 13 2003 - 17:06:46 CDT

Re: mbrc (was: buffer cache - once again)