Re: Buffer Cache performance

From: Steve Adams <steve.adams_at_ixora.com.au>
Date: Mon, 01 Nov 1999 00:25:51 GMT
Message-ID: <381cda9b.20260442@news.eagles.bbs.net.au>

Hi All,

When big tables are scanned, the blocks are read in the MRU end of the cache if the CACHE directive has been specified, but ONLY until such time as the number of blocks specified by the cache_size_threshold parameter have been read. Thereafter, further blocks will be read into the LRU end, regardless of the CACHE directive.

The cache_size_threshold parameter defaults to 10% of db_block_buffers. If you use the CACHE directive, it is wise to check routinely that those tables have not grown beyond this threshold.

Regards,
Steve Adams

http://www.ixora.com.au/

http://www.oreilly.com/catalog/orinternals/

http://www.christianity.com.au/

On Sun, 31 Oct 1999 18:02:39 -0000, "terry norman" <scott_at_tiger9.freeserve.co.uk> wrote:

>Julio,
>
>We had a problem with a big table which was set to "cache" and not
>"nocache", taking over the buffers, or at least a lot of them, whenever it
>was scanned. Increasing the value of db_block_buffers brought little
>increase in hit rate because of this. So it might be worth having a look to
>see if you've got any big cached tables. Just a thought.
>
>Regards,
>Terry.
Received on Sun Oct 31 1999 - 18:25:51 CST