You might try doing it in batches. It could be that you have a lot of
overhead if you call update or insert for every row and you're not in
a batch.

Alternatively, you might be able to do the work in all SQL.  This
might be a much better approach.

You haven't provided enough detail to suggest a better approach. My only
thought is that 57,000 records might be able to fit into memory easily,
in which case you could read the whole thing into memory quickly, do
your calcs there, and then dump it all back to the database.

My guess is that if this process is currently taking 8 hours with such a
small amount of data, you're probably doing very inefficient database
access. So get it out of the database altogether.

Whether this will work or not depends on your app, of course.

My only other thought: temporary tables are your friend.

Ankit <shahankit@gmail.com> wrote in news:1191463557.035295.92260@
50g2000hsm.googlegroups.com:

Have you analyzed the performance of the query?  EXPLAIN PLAN is your friend,
as are the more sophisticated analysis tools available in Oracle.

How long does the same query, or set of queries take when run outside of
Java, in SQL*Plus?  You need to determine if it is database performance that
is your bottleneck or if it is java performance that is your bottleneck.

If you are doing calculation or processing (in Java)  of the rows returned by
your queries on these two tables, then you should perhaps also look at
splitting up the result sets from the two queries into chunks and processing
multiple chunks in multiple threads in your program.

Without knowing the structure of your database schema (tables and indexes)
and the specifics of your queries, it would be very difficult to advise you
further, though.

Cheers!

BTW: did you know that most DBs will let you do a lot of different
calculations (including date/time) in SQL?