FreeTDS Development Group

[Frediano Ziglio <freddy77 AT gmail.com>]

Il giorno sab, 03/05/2008 alle 08.51 +0200, Sebastien FLAESCH ha
scritto:
> James K. Lowden wrote:
> > Rajesh Akolu wrote:
> >> We are using FreeTDS 0.64 (TDS version 8.0) and unixODBC 2.2.12 on RH
> >> Linux 3.4.6 and connecting to Microsoft SQL Server 2000 from Linux. 
> > ...
> >> We are looking for parameters which can increase the performance
> > 
> > You're looking for magic pixie dust to "increase performance" while
> > providing no evidence for the notion that performance can be increased. 
> > You seem to assume it can be, but why?  In a database application,
> > bottlenecks don't show up in the network layer until the server and client
> > are running at top speed.  
> > 
> > You can't manage what you don't measure.  Instead of asking the Internet
> > for random tuning suggestions, monitor your application and discover where
> > the delays are.  Then address yourself to those areas.  You'll have the
> > satisfaction of learning something and doing the job right.  And,
> > possibly, of increasing performance.  
> > 
> > --jkl
> > _______________________________________________
> > FreeTDS mailing list
> > FreeTDS AT lists.ibiblio.org
> > http://lists.ibiblio.org/mailman/listinfo/freetds
> 
> Rajesh,
> 
> With the native MDAC and SNAC drivers of SQL Server, we got very good
> performance improvements by using array fetch (SQL_ATTR_ROW_ARRAY_SIZE).
> 
> I am not quite sure if using array fetch makes a difference with FreeTDS,
> but so far we use it as it was really efficient with the native drivers.
> 
> Note that we are not using the default cursor: we use server-side cursors.
> 
> Cheers,
> Seb

Yes, I can confirm that this works even for FreeTDS. The reason is that
you do a single server request for N rows. For normal cursors all rows
are returned from server but with server side cursors you require
row_array_size rows at a time (by default 1), this is the cause of
improve, less round-trips!

freddy77