Importing data, we have the following speeds (in ms)


I had assumed that actually creating the objects would be the slow part, but it appears that the flush to database is the bigger bottleneck. Somehow I find it hard to believe that the database is the problem unless its a combination of jdbc and statement creation? Are the insert/update statements generated in batches, or send 1-at-a-time to the database?

Are there any optimization tricks I should be aware of (on the hibernate side)?

Using: postrgreSQL, linux, and jboss 3.2.1 w/tomcat

Thnx.

LOL, you have not been working with databases very long, huh?

During flush() is when all the work Hibernate does synchronizing in-memory state with the database. So this typically is the slowest phase

actually I've been working with databases for a long time, but when postgres does 10000+ inserts in a minute from a psql command (batch file), it seems a bit funny that it would be that much slower.. thats why I expect it to be the jdbc that slows things down, rather than the netowrk speed since similar timings are seen when running both DB and jboss/hibernate on the same machine.

Cheers..

While I have your attention, any idea how to catch a jboss transaction timeout from hibernate? session.flush() just goes on until all have been flushed, even if a timeout occurred in the first minute or so.. and doesn't seem to notice any errors when it returns from the call. What am i missing? how can it keep sending the statements if the connection is not available?

thnx...

No idea. I don't use JBoss.

Note that the postgres JDBC driver is incredibly slow, so we expect inserts done from Java to be much slower than inserts done from a batch file.

Aha! That would do it. I was unawares that postgres's jdbc was known for being slow as molasses.... thnx.
Ati.