I followed an example in the tutorial/book to the letter, scrolling with the cursor, flushing, clearing the context.
It has worked; however, the sql log showed updates as separate statements. I went further to profile the database (sql server).
The profiler again showed that execution happened in a non-batching manner.

I tried to debug Hibernate source code, but could see that when the batching is called for
in org.hibernate.engine.ActionQueue.executeActions(List) with
session.getBatcher().executeBatch() -> org.hibernate.jdbc.AbstractBatcher.executeBatch(),
internal batchUpdate (PreparedStatement) is null, preventing batch execution.

I have been looking around for a long time, but failed to find what needs to happen for batch to work.
I did set hibernate.jdbc.batch.size in persistence.xml and hibernate.cfg.xml files.
I am using annotations if that helps.

Please help! I am trying to move our company towards Hibernate.
Thanks in advance.

Turns out Hibernate does batching only if the app is not using optimistic locking OR
(the entity is not versioned && use optimistic locking) OR the property hibernate.jdbc.batch_versioned_data
is set to true.

This can be seen in the code here org.hibernate.persister.entity.AbstractEntityPersister.isBatchable()
My code had @Version on a column/

SQL Server execution profile does not seem to differ though: statement declaration followed by N executions.
Though, batching does seem to make run time for the update faster.

Another way to batch updates/inserts, etc is with session.doWork().

Hi,

I did make some experience with Hibernate & Batch-updates on SQLServer2008.
It worked to me already with the first try (I use a @Version column in each entity class),
anyway the performance gain was not very remerkable:
It was ca. 10% faster than without batching (I must mention here that we have a fast network).

I used

on large transaction where I do perform more than 1000 insert and updates per flush or commit.

If there raises a SQLException during batch-updates, then the message is not so clear and precise as without batching.
Due this disadvantage I personally will use batching only during data-migration but not afterwards in the live-system.