Re: Performance issue while loading large amount of data

From: Richard Foote <richard.foote_at_bigpond.com>
Date: Thu, 16 Jan 2003 23:32:24 +1000
Message-ID: <PmxV9.25564$jM5.67468@newsfeeds.bigpond.com>

"Noons" <nsouto_at_optusnet.com.au.nospam> wrote in message news:Xns9305BF2870F1mineminemine_at_210.49.20.254... <snip>
>
> This would get much worse if the table/index was not empty before,
> I'd expect? Because then it might be recording b-tree branches
> and splits etc of the already existing index blocks, depending
> on what values of keys we were inserting. Correct?

Hi Nuno,

Potentially yes.

>
>
> >
> > I now carry on and insert 340 rows into this table.
>
>
> Hope you really got that into a script! :D

Sore fingers ;)

<snip>

> What I find surprising here is the header being saved so many
> times. In the same transaction I can't see why that would
> be needed? I mean, it's not like we'd "partially" rollback, no?
> Just saving the original block's header once would do the job to
> reset the block in case of rollback? Or is it in case we
> have savepoints?

It's the *undo record header* that's repeated which contains "overhead" details in relation to what the undo is for (block address info, no of rows, etc).

<snip>
> Thanks a lot for this, Richard. If any1 posts your
> Wednesday photo somewhere, let me know and I'll send
> "the boys" around to straighten them up! :D

My pleasure !!

Cheers

Richard Received on Thu Jan 16 2003 - 07:32:24 CST