Re: Run Oracle without transactions?

Just a little follow up -
So this is what "redo wastage" refers to in the bstat/estat report?

Thanks,
- Dc.

Jonathan Lewis wrote:

> Every time you commit LGWR has to write a complete
> number of O/S blocks to disc. On average every write
> of LGWR wastes half an operating system block.
>
> If you commit every row, you may find that each
> write of LGWR is only a couple of blocks, which
> tends to be about 25% wastage (half a block out of
> every 2), but if you commit every couple of hundred
> rows you may find that LGWR writes about 50
> blocks per write request, with wastage down to
> 1% (half a block in 50).
>
> I used this to good effect recently in fact - one site I
> visited was generate 125Mb of redo writes in 10 seconds,
> of which 25Mb was wastage. When they changed
> to committing every 500 rows the total redo DROPPED
> to 113MB in 10 seconds, of which only 3Mb was wastage.
> Net result - they were processing 10% more data in the time
> (actual redo had gone up from 100Mb to 110 Mb - and this
> was a fair measure of work done).
>
> --
>
> Jonathan Lewis
> Yet another Oracle-related web site: www.jlcomp.demon.co.uk
>
> Doug Cowles wrote in message <37883818.5BC94C58_at_bigfoot.com>...
> >Just curious, why is that going to drive up redo log wastage.
> >One extra byte or two for every commit?
> >Just wondering..
Received on Mon Jul 12 1999 - 16:04:21 CDT