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: Import performance.

Re: Import performance.

From: Petter Hansen <pettehan_at_online.no>
Date: Tue, 27 Oct 1998 23:21:04 +0100
Message-ID: <36364750.82522D37@online.no>


hmm.

commit=n is better. (Demands a big rollback segment)

Douglas Dunyan wrote:

> On large imports, I found that turning off archive mode will save the
> most time.
>
> Here's the trick, if you have the luxury.
>
> We run everything redundant, including disks. Before attempting the
> import, put in a third shadow set, and turn off archiving. When done
> with the import, do a normal shutdown, and break the 3rd shadow set
> (you can use that for cold backup!) and then turn archive mode back
> on....
>
> With large imports, most of the performance hit is i/o due to archive
> mode.
>
> Doug
>
> Linda Montgomery wrote:
> >
> > There are several changes we make to increase import speeds. Here is our little
> > checklist of things to check/change before a big import. Afterwards we put
> > everything back like it was.
> > 1. Make sure redo logs are large enough & that there are enough groups. Try to
> > keep log switches at least 3 minutes or more apart. We have been increasing
> > logfile size to 30M minumum ( a size that worked for our environment) if
> > smaller than that in production.
> > 2. Place each logfile on a seperate drive (or ping pong between drives) to
> > eliminate contention.
> > 3. Set db writers = # of disk drives * 2
> > 4. Set init.ora parameter checkpoint_process = true
> > 5. Verify log_simultaneous_copies is corret for the # of processors, or remove
> > parameter if on a single cpu machine.
> > 6. Check for adequete rollback & temp tablespace sizes & that their initial &
> > next extent will allow them to grow to the maximum value.
> > 7.Be sure to use commit=y for large tables
> > 8. Increase size of init.ora parameter log_buffer to 1,024,000
> >
> > Linda
> > montgolh_at_jmu.edu
> >
> > MotoX wrote:
> >
> > > Correct, only EXPORT supports DIRECT.
> > >
> > > And remember to take a *backup* after recreation of the indexes, unless you
> > > want to go through the whole procedure again should you suffer media
> > > failure...
> > >
> > > MotoX.
> > >
> > > Murali Kazhipurath wrote in message <362d879c.0_at_blushng.jps.net>...
> > > >From what I have understood, DIRECT PATH only works with EXPORT. Here are a
> > > >few tips:
> > > >
> > > >1. Drop all Indexes.
> > > >2. Import without indexes using the option indexes=n.
> > > >3. Rebuild indexes using the UNRECOVERABLE option.
> > > >
> > > >Hope this helps.
> > > >
> > > >Murali
> > > >murali_at_jps.net
> > > >
> > > >
> > > >xsun1913_at_my-dejanews.com wrote in message
> > > ><70ilit$2j0$1_at_nnrp1.dejanews.com>...
> > > >>In article <01bdf5d0$fabc77a0$5feaabc3_at_ken-bailey>,
> > > >> "Ken Bailey" <Ken.Bailey_at_BLAENAU_GWENT.btinternet.com> wrote:
> > > >>> Can anyone give me any guidelines on improving import performace (
> > > >>> parameters etc) ?
> > > >>>
> > > >>>
> > > >>
> > > >>Use DIRECT PATH when running import; Set buffer to a large number.
> > > >>
> > > >>
> > > >>Xuening Sun
> > > >>
> > > >>
> > > >>
> > > >>-----------== Posted via Deja News, The Discussion Network ==----------
> > > >>http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
> > > >
> > > >
Received on Tue Oct 27 1998 - 16:21:04 CST

Original text of this message

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