Tom Lane wrote:
It sounds like you're incurring a network round trip for each row, which will be expensive even for a co-located application. Perhaps Oracle's JDBC driver is smart enough to avoid that. I'm not sure what tricks are available for bulk loading with our JDBC driver --- the page Heikki mentioned explains things from a server perspective but I dunno how that translates into JDBC. The folks who hang out on pgsql-jdbc could probably give you some tips.
OP said he's using addBatch/executeBatch with a batch size of 100. The JDBC driver sends the whole batch before waiting for responses.
-- Heikki Linnakangas EnterpriseDB http://www.enterprisedb.com
