Re: Buffer cache statistics (ratios) and CBO SQL optimization?

On Mon, 29 Dec 2003 10:10:01 -0000, "Niall Litchfield" <n-litchfield_at_audit-commission.gov.uk> wrote:

>I'd be a little surprised if they argued that the two were almost equally
>expensive, I'm pretty sure that the figure is something like 20-100 times
>which is significantly smaller than the figures commonly given but not
>almost equal.

I really do not want to get into semantics. The gist of the story was you shouldn't think a LIO doesn't cost anything. (Maybe I should just try to re-find it. It could be it included measurements comparing the duration of logical IO vs physical IO). Hence you shouldn't focus on PIO alone. This usually results in 'throwing more memory at the problem', which usually doesn't work at all. On an interesting side-note, I once administrated a database, where all the init.ora parameters were still on their initial defaults. The BCHR was about *20* percent. Guess what? No one complained. Surely at that time I bumped up the buffer cache, but I didn't get any positive comments.

--
Sybrand Bakker, Senior Oracle DBA