Hi,

I tried to compare JDBC versus Hibernate using the performance test framework provided by the OJB project : (http://db.apache.org/ojb/performance.html#OtherTools).

I wrote HibernatePerfTest and HibernatePerfHandle classes, and changed a bit the JDBCPerfTest/Handle classes provided.

I changed the following things in the JDBC classes :

* isolate this class from any OJB Broker or Identity class

* use the auto-increment feature of mysql to have the same primary key generator for both tests

I use mysql 4.0.16. on a server.

I get the following results which I don't understand :

09:45:57,015 INFO SessionFactoryImpl:272 - Query language substitutions: {}
allocated memory after test: 23752 kb


========================================================
OJB PERFORMANCE TEST SUMMARY
10 concurrent threads, handle 2000 objects per thread
- performance mode
========================================================
API Period Total Insert Fetch Update Delete
[sec] [sec] [msec] [msec] [msec] [msec]
--------------------------------------------------------
JDBC 24.177 22.832 7322 187 8580 6743
Hiber 30.702 29.891 13438 214 8952 7286
========================================================



Why are the inserts so slow, what did i do wrong with my test set up, i am sure i am missing something, so if you can help me to get closer to the less than 10% overhead claimed...

Hibernate normally beats OJB comfortably in their own performance tests so I'm not sure what you did wrong. Probably just a basic error in you tests.


P.S. for fairness you should compare Hibernate vs. their ODMG API since the PB API does not offer full object persistence semantics. In this case we win by a really large margin.

oh in this case you are comparing vs. direct JDBC?

well, still not sure what you did wrong but I guarantee that overhead on inserts is tiny.

I don't know anything about the OJB tests, but using the getGeneratedKeys patch I submitted against DB2 running on a remote iSeries server I get close to double the performance for an insert with generated keys.

Ah possibly OJB is using getGeneratedKeys().

We will integrate this patch. I'm just not sure if we do it before 2.1 final, or wait for 2.1.1.

I downloaded OJB and it doesn't use getGeneratedKeys so I can't explain the difference. If I get some time, I may try a test of my own. There is no real rush to add the patch, I'm getting pretty good at merging but it is a free performance boost.