At 06:48 03/06/06 +0200, you wrote:

>On Friday 06 June 2003 05:11, Almond wrote:
> > Somebody ask about the distribution, my consideration is the fs
> > performance. We have reiserFS, XFS, JFS and ext3. I have an application
> > which have a db about 5~10G size. Have anybody done some bench marking on
> > fs vs firebird.
>
>I haven't seen any benchmarks, but they would be interesting, wouldn't they?
>
>Common sense indicates that reiserFS should be slowest and XFS fastest. In
>the
>absence of any hard facts I'd stick with ext3. Speed is great, but data
>integrity is better still. ext3 seems to be a good compromise.

Is ext3 good ? I have some bad experience with it. It have bug on smp
machine not long ago.

> > 1. Difference on pre-allocation of spaces. Some reading mentioned that this
> > would enhance the performance since the db engine need not extend the file
> > through the fs which is very time consuming (?).
>
>Pre-allocating space is in itself time consuming. All you are doing is time
>shifting. If you do try pre-allocating and you use dummy data don't forget to
>drop the table. (ie, don't use 'delete from' to get rid of the data)
>Otherwise you'll lose the time saved to garbage collection.

Yes, it is time consuming. But only once. If the time saved in daily
operation is great, I'll give it a try. What I think is: create a dummy db
with a dummy table (huge!), gzip it. Every time when I need a new db, I
unzip the file and drop the dummy table, and hope the daily transaction
time would reduced. Is this idea useful ?

> > 2. fb page size vs fs block size. Is there any optimization rules for
> > difference size of db ?
>
>Disc i/o is independant of page size.

What is the rule of thumb for page size ? Is 4K the magic number ?

>Paul
>--
>Paul Reeves
>http://www.ibphoenix.com
>Supporting users of Firebird and InterBase
>
>
>
>To unsubscribe from this group, send an email to:
>ib-support-unsubscribe@egroups.com
>
>
>
>Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/