We experiment low performance with SQL-server 2008 when performing massive INSERT queries. These queries are enclosed into transactions by batches of 25, 50, 100, or 200 rows.
Different factors influences the DB performance:
the number of concurrent queries handle at the same time by the database
less obviously, the complexity of the INSERT (normal speed when 8 columns, but 60 times slower when 20 columns !)
and more unexpectedly, the client application: we have tested jTDS, SQSH (command line) and Toad (GUI).
The following graph shows how Toad is much faster and steady than the 2 other applications:
NB: we have also tested SQL Server Express and it's as fast as Toad.
How can Toad be more performance than the two other lighter application? How can we reproduce this performance with jTDS or SQSH? Any idea?
We offer a beer !
François