Re: db block size, too big wasting buffer?

Keep in mind that this is one sure-fire route to performance disaster.

Multiple block sizes were invented to enable transportable tablespaces to work properly (ie, transport between databases that were created with different block sizes). No doubt you can use them to do as you describe, but it will generally be an utterly inappropriate thing to do.

There is no keep or recycle pool possible for a non-standard block size, so there's your first big performance penalty. SYSTEM and all temporary tablespaces must always be of the standard block size -so there's your next performance hit if you get that decision wrong.

It is flabby DBAing of the dangerous kind to think that multiple block sizes is going to be of performance assistance. On a file-system based database, there is still a 'right' block size, and anything other than that will bring a performance hit.

Regards
HJR "John Beresniewicz" <jberesniewicz_at_comcast.net> wrote in message news:D3NR8.375033$%y.28504063_at_bin4.nnrp.aus1.giganews.com...
> Keep in mind that on 9i you can could use a 16K default block size as well
> as a dedicated 4K cache (block) size just for the OLTP table. This should
> minimize block contention as well as isolate (and tune) buffer cache and
> provide dedicated LRU working sets for this presumably all-important
object.
>
> Sure sounds like having your cake and eating it too in this case.
>
>
>
Received on Tue Jun 25 2002 - 03:33:37 CDT