Hi:
I am in the process of evaluating the use of Hibernate for a call center application that would have close to 250 concurrent users doing mainly select operations (quering information) on close to 100 tables.

The expected no. of records in the db is likely to be 100 Million.

We have decided to use Struts for the Presentation layer of our application.

We've developed a small PoC for Hibernate on some of the most complicated scenarios we envision in the application and are convinced that it works well. However, there is no way for us to validate whether Hibernate is scalable enough to support 250 odd concurrent users and if so what is the performance impact it would have. Our biggest requirement is on the Performance front considering it's a call center application.

An early response and help from all of you who have used this would be of utmost importance.

Thanks.

Its better to measure in terms of max concurrent requests than concurrent users. Also, it is all relative to how much hardware you plan to throw at the problem.

However, 200 concurrent users is not a particularly large number. I couldn't image that would be a problem.

Gavin: That was fast!

Could you let us know how much of hardware would we normally require to support 250 odd concurrent users? We are thinking of implementing this initially on Oracle but moving onto other Dbs depending on our clients needs.

Any insights would be helpful.

we have tested 500 concurrent request with loadrunner on 2 tomcat with sticky session.
The db (oracle) was having problems before the tomcats...
we were hitting 600 records once (3 linked tables) per requests without cache.

I am pretty new with Hibernate, but I have been working with Oracle for 6 years. Oracle can handle many concurrent users. Some of our clients have 600 - 1200 users(not running hibernate). I did some benchmarking at about 3 years ago at Sun Microsystems on Sun E450's with 4x480Mhz Ultra Sparc II CPU's and 4 GB memory with Oracle 8i and we able to get up 750 users.

This totally depends on if the database/application has been tuned properly. For example, if you have a table in which you know will eventually have a large number of rows, consider creating a Partioned Table.

I highly recommend hiring a experienced Oracle DBA.

>>I did some benchmarking at about 3 years ago at Sun Microsystems on Sun E450's with 4x480Mhz Ultra Sparc II CPU's and 4 GB memory with Oracle 8i and we able to get up 750 users. <<

Did you used some kind of standard benchmark transaction / scenario? There is everytime a problem by stating numbers without knowing what they represent.

Also I would like to add, while DB optimization is important, it should not be the base of a decision on wether to use hibernate or not. Just create some kind of prototypes imitating the main task of the real application. Implement it with diffrent persitence 'layers' and you will have a good answer to the question.

But you know, Hibernate is a lightweight persistence framework. This means, if there is a limitation in the concurrent servable user amount, the limitation is likely to be also present implementing the app using JDBC / DAO for instance.