Re: Disable Buffering

From: DA Morgan <damorgan_at_exesolutions.com>
Date: Wed, 26 Feb 2003 14:00:52 -0800
Message-ID: <3E5D3914.69267779@exesolutions.com>

"Jörg Brenninkmeyer" wrote:

> "Sybrand Bakker" <gooiditweg_at_sybrandb.demon.nl> schrieb im Newsbeitrag
> news:os0q5v01buufi34o3c24sp0jau1bkicg4v_at_4ax.com...
> > On Wed, 26 Feb 2003 17:36:10 +0100, "Jörg Brenninkmeyer"
> > <jbrenni_at_uni-muenster.de> wrote:
> >
> > > But since I want to do some queries which are
> > >based on the same table _independently_ of each other, I must at least
> empty
> > >/ override the cache between those queries. Is there a way to do this
> > >automatically? Or do I have to do some queries on a completely different
> > >table?
> > >
> > >Thanks,
> > > Jörg
> > >
> >
> > You REALLY don't know what you are talking about, do you?
> >
> > Oracle has READ CONSISTENCY. Know what that means?
> > EVERY QUERY will run INDEPENDENTLY!!!!!!!!!!!

>

> No need to get unfriendly.

>

> If I do a first query, the tables needed are loaded in the cache, if it's
> big enough. If I then run a second query, which is different, but based on
> the same tables, it can be executed just using the data which is cached from
> the first query. SO, the queries are NOT independent from each other
> REGARDING CACHING. And this is why I want to flush the cash inbetween those
> two queries. In the tests reading from disk SHOULD be an influencing factor,
> and since real data would be much more, it's quite realistic that the tables
> needed for the next query would NOT be cached for the query executed before.

What you are suggesting would only make sense if you were not on a multiuser system running mulitple transactions between shutdowns.

Reality is that the tables most likely to be used will mostl likely be in memory (assuming a decent design) and testing after flushing the cache will give you numbers that will only be valid after a cold startup for the first five or six queries.

Why test under totally unrealistic conditions?

Daniel Morgan Received on Wed Feb 26 2003 - 16:00:52 CST