> > > i have a table with only 1 record. i run updates like "update
> > > field=field+? where pk_id=?" on this record. the updates are
> > > executed by one process, and come in transactions which include
> > > about 1000 updates. while the first transaction takes about
> 1.5sec
> > > the 100. transactions takes up to 15sec. and it gets slower and
> > > slower. what could be the reason for this ? my first guess was
> the
> > > mga , but there are no concurrent transactions, since the
> statements
> > > are executed sequentially by just one process. does anybody know
> > > what is causing this decrease in performance?
> >
> > Do you commit the transaction?
> >
> > Do you have a read transaction opened?
>
> i commit the transaction every 1000 updates. there are no other
> transactions running against this database.

No SELECTs?

> > Take a look at the transaction counters with "gstat" and post
> > them here.
> >
>
> at what time should i check the counters? before i start the
> transactions, during the run of the program, or afterwards?

When it's slow.

With regards,

Martijn Tonies
Database Workbench - developer tool for InterBase, Firebird, MySQL & MS SQL
Server.
Upscene Productions
http://www.upscene.com