Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Run Oracle without transactions?

Re: Run Oracle without transactions?

From: Doug Cowles <dcowles_at_bigfoot.com>
Date: Sun, 11 Jul 1999 02:22:17 -0400
Message-ID: <37883818.5BC94C58@bigfoot.com>


Just curious, why is that going to drive up redo log wastage. One extra byte or two for every commit? Just wondering..
- Dc.

Jonathan Lewis wrote:

> I suspect that there is no cheap fix for your
> problem, too many black-boxes with the
> potential for complicated twiddly bits.
>
> You ought to have an option to commit
> one object at a time within your loading
> front-end. This will save the massive
> rollback, though it could drive the redo
> log wastage (hence I/O) up quite a bit.
> (If there isn't an option to commit for
> each object, there should be).
>
> You could look at the option for producing
> a second loader, using common code to
> generate flat files corresponding to the
> relational tables that would be loaded, and
> then use a fast loader to pump the flat
> files into the database.
>
> To get an insight into what the black boxes
> are doing, you could send a sql_trace call
> to the process that gets the data into the
> database - it might give you some clues
> on a least expensive strategy.
>
> --
>
> Jonathan Lewis
> Yet another Oracle-related web site: www.jlcomp.demon.co.uk
>
> Anita Krishnamurthy wrote in message <37840543.F914E81F_at_cls.uob.com.sg>...
> >We are using Oracle 8.
> >
> >Ours' is an object oriented system, and we are trying to load objects into
> >the database. We have another utitlity which does the mapping for us, from
> >objects to relational database(which means we do not have control over
> which
> >tables a single object maps to, or where the data is stored). This layer
> uses
> >SQL statements, and we have to go through this layer. And for the same
> reason
> >that we are loading objects, we cannot define a logical commit point, since
> >we do not know at which point the object references are resolved
> completely.
Received on Sun Jul 11 1999 - 01:22:17 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US