Also you are limited by how fast IO can be processed on the disk device(s).
A dick cache can temporarily help on peaks, but not on sustained throughput.
You may need to increase your raw available IO service rate. (more disks)
But try the advice below first.
Jim
<jdarrah_co_at_my-deja.com> wrote in message
news:8slbmt$oki$1_at_nnrp1.deja.com...
>
>
> If the secondary inserts you're doing are for multiple rows, you can
> use the APPEND hint. Example:
> INSERT /*+ APPEND */ into a (select * from b)
>
> This will direct oracle to insert rows above the high water mark rather
> that blocks that are on the freelist. This works well if you are
> inserting serveral rows into a table at a time in batch. Also,
> altering the table NOLOGGING can help with the inserts as well. If the
> target table in question has several indexes and you don't need them
> all for your load, it may be quicker to drop, load, and then re-create
> the index.
>
> In article <ROcH5.1093$G3.179425280_at_news.telia.no>,
> "Randi Wølner" <rw_at_computas.no> wrote:
> > I am working with conversion of a large quantity of data that are
first read
> > from text files into Oracle tables, and then processed and put into
> > different tables. I was very satisfied with using the 'Direct Path'
option
> > of SQL*Loader, so reading the data into Oracle goes real fast.
> >
> > But - after loading the data I do lots of inserts and updates to put
the
> > data into many tables, and this takes lots of time. When this job
will be
> > done 'for real' I will have the database on my own.
> > Is there any option in Oracle I could use to 'disable' the use of
rollback
> > segments, to make things speed up?
> >
> > Thanks,
> > Randi Wølner
> >
> >
>
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
Received on Thu Oct 19 2000 - 00:31:04 CDT
