Re: 9i multi cache buffer

> > > Hi Howard,
> > >
> > > OS Platform => Solaris, with raw devices.
> > >
> >
> > 16K.
>
> Hi Howard,
>
> 16K is the best block size here, period.
>

Richard, if you want me to revise my previous reply where I said I only needed to ask about the O/S then I shall: by "O/S" I of course meant "O/S and the file system it imples". You have an O/S but don't have a file system. The sky is therefore the limit, and you are free to choose whatever block size you fancy. I would still go for 16K because contention issues would be manageable by other means, and yet it's big enough to get the benefit of good I/O on the index.

So, yes. Period.

> You can think of no situation or set of conditions where another block
size
> might be more appropriate ? .
>

On the whole, no.

HJR
> Cheers
>
> Richard
>
> >
> > Regards
> > HJR
> >
> >
> > > Correct block size is ...
> > >
> > > Cheers
> > >
> > > Richard
> > >
> > > >
> > > > ;-)
> > > >
> > > >
> > > >
> > > > > or
> > > > >
> > > > > b) It depends
> > > > >
> > > >
> > > > ...on the O/S platform.
> > > >
> > > > > or (failing everything else)
> > > > >
> > > > > 8K.
> > > >
> > > > Hmmm. It will do. Unless you're on AIX or Linux or NT.
> > > >
> > > > :-)
> > > > HJR
> > > >
> > > >
> > > > > Over-simplification, I know, but it answers the question.
> > > > >
> > > > > In any case, in real life it's usually the crappy SQL that makes
the
> > > > > difference.
> > > > >
> > > > > Regards,
> > > > > Paul
> > > > >
> > > > >
> > > > >
> > > >
> > > >
> > >
> > >
> >
> >
>
>
Received on Mon Nov 18 2002 - 17:34:42 CST