Hi, I am Fernando:

I have detected a performance issue and i need to kmow if there is any solution.

I am making stress testing with different data base enviroments. All enviroments are is same scenario. My results are as follows:

SQL-SERVER 2000-> de 10 a 100 workers -> 1800 tps.
ORACLE 9i -> de 10 a 100 workers -> only 15 tps.
SYBASE ASE -> de 10 a 100 workers -> only 21 tps.

In each one, i have used their specific dialects (MsSql2000Dialect, Oracle9Dialect and SybaseDialect).

After that i have chaged, only for Sybase the property "Hibermate.Dialect" and then the stress testing has grown to 900 tps, a success!. But naturally, i will not be able to use the same for Oracle because it will not generate properly the mapping (for example it would fail in ORACLE used .SetMaxResults()).

Is there any other way to increase the performance?
Is the dialect really the problem?

Regards.

Could you post your test code please? It is really hard to tell what causes the behavior you have. I would be interested in looking into this in our lab.

Hi Konstantin:

I have sent you a mail.

Thanks, for your interest:

The stress test is used with the Microsoft`s Framework, but with some changes.

These tests have been done, first with sql server 2000 environment, Oracle9i, Oracle RDB, Sybase ASE, Sybase IQ, using both native drivers and ODBC, and the average in each one of the environment has been 800 and 2000 tps. All with “command”

After I have tried to do the same test with Nhibermate, being the result before.

I add the sql to insert. Only is neccesary to do the test of read.

The reading have been done with a million of registers in each enviroment..

I add the code of the test for sql. For Sybase is exactly the same, but changing the .Config with the Sybase parameters. I have used ODBC, for Sybase ASE because is the best result that I obtained in the previous stress test.

Now i am ill. My knee is damage. That´s why I am responding solate.

Whatever doubt you have, ask me.

Best wishes and thanks.

Fernando Colmenarejo.


For the moment, we could strat with SQL and SYBASE.




HARDWARE BBDD SYBASE AND SQL SERVER:

Sistema Operativo:

Windows 2000 SP4

Procesador / es:

4 Procesadores Intel Xeon a 3,2 GHz

RAM:

4 GB

HD:

400 GB en una unidad lógica de una SAN en cluster

The problem to Sybase is that not use "Top" to limit the rows in the SQL. SybaseDialect not use TOP for it, the time of the database is very long, because Nhibermate limits the number of rows after get in all. It´s the first problem because you would use SQL and not the SetMaxResults().

how can i optimize the criteria without using SQL?

sorry by mi english, I only read English.

Fernando Colmenarejo