Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: cbc latch assignment on db_cache resize

Re: cbc latch assignment on db_cache resize

From: Jonathan Lewis <jonathan_at_jlcomp.demon.co.uk>
Date: Thu, 18 Jul 2002 12:47:52 +0100
Message-ID: <1026992834.23504.1.nnrp-01.9e984b29@news.demon.co.uk>

Just did a couple more experiments -
the first hypothesis is wrong.

I added a 2k cache sized at 4M in a
96MB limit, and saw 512 latches jump
to 1024 latches -- a reasonable response according to my theory.

But my test with 512MB is not repeatable. (I have to say I was a bit surprised to see 1536 as I was expecting a 1024 or 2048,
not halfway in between).

However the nail in the coffin was:
512MB sga_size
Start at 4M for a 2K cache.
1024 latches allocated on startup.

Increase the 2K cache dynamically
to 320MB - no change in latch count.

Restart the system with 320MB of 2k cache, 2049 latches allocated on startup (as expected using Steve Adams' formula for latch count).

Ho-hum.,

--
Jonathan Lewis
http://www.jlcomp.demon.co.uk

Next Seminars
 UK  July / Sept
 Australia August
 Malaysia September
 USA (MI) November

http://www.jlcomp.demon.co.uk/seminar.html

Jonathan Lewis wrote in message
<1026990884.22620.0.nnrp-01.9e984b29_at_news.demon.co.uk>...

>Andrew,
>After your first email, I did a couple of quick tests
>on an NT 9.2.
>
>I happen to have an 8K database with 4K blocks
>Starting with
> sga_max_size = 256Mb
> db_cache_size = 8M
> db_4K_cache-size = 8M
>I see 1024 cbc latches.
>
>If I restart the database, but change the
>sga_max_size to 512MB, I get 1536 cbc
>latches.
>
>So I was going to test the theory that the
>number of cbc latches is determined by
>largest number of buffers you could demand
>given the smallest block size with a defined
>cache.
>
>BTW - the number of latches in 9.2 seems
>to revolve around significant powers to 2.
>I think you are thinking of the number of
>hash buckets when you say 'prime number
>close to twice the buffers'.
>
>
>--
>Jonathan Lewis
>http://www.jlcomp.demon.co.uk
>
>Next Seminars
>UK July / Sept
>Australia August
>Malaysia September
>USA (MI) November
>
>http://www.jlcomp.demon.co.uk/seminar.html
>
>Andrew Mobbs wrote in message ...
>>
>>Having now had a chance to play with this, it appears that, at least,
>>the values for _db_block_hash_latches and _db_block_hash_buckets in
>>x$ksppcv stay constant when db_cache_size changes.
>>
>>This, unfortunately for me, means that shrinking the SGA shouldn't make
>>cache buffers chains contention worse. However, should we then be wary
>>of starting with a small SGA and dynamically growing it to something
>>quite large?
>>
>>Interestingly, 9iR2 (at least, maybe 9.0.1 too), _db_block_hash_latches
>>has gone back to being prime but now is a prime "near" (not sure exactly
>>how it's chosen) twice the number of block buffers.
>>
>>--
>>Andrew Mobbs - http://www.chiark.greenend.org.uk/~andrewm/
>
>
>
>
Received on Thu Jul 18 2002 - 06:47:52 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US