Re: sizing buffer cache from "scratch"?

Fredrik wrote:

> Hi group!
>
> If we want to take a more scientific approach - just boldly chosing a
> round figure and watch hit ratio seems somewhat too random... How would
> you go about when trying to find/estimate a reasonable number of buffers
> for the db block buffer cache? Is there some "well-practiced" route to
> identify those bound to be most used blocks?
>
> Maybe I should add that this is from a standpoint with pre-packaged
> systems that doesn't come with much knowledge of data and sql-code (but
> instance and/or storage parameters are often default or something
> worse).
>
> Your replies are most appreciated.
> Regards
> Fredrik
>
>
>

Hi Fredrik

You are asking a general question about a 'general' system without telling anything (mem-size, no. cpu's and type....), an 'general' application without number of expected users, type etc.

So if I could give you the figures - just like that - I can also predict the future and make gold.

With the massive of info you have given, I predict - use the memory you got, and use it well.

/Svend Received on Tue Jan 08 2002 - 14:20:35 CST