[JBoss-user] Caching large data structures

[Davide]

Hi,

My staff is studing the porting of an online trading proprietary application 
with J2EE technology and JBoss. This application handles about 60.000 Tx per 
day. We need a cluster for fail over and load balancing scaling capabilities.

We made some test on a sigle JBoss instance using commit option A on entity 
beans obtaining encouraging results for a single PIV machine.
Then we setup a clustered two JBoss instances setting commit option B on our 
entity beans. In spite of two machine handling, each one, half of the load, 
performances has noticeably decreased.
The database was on a third machine on both the tests.

This application need to cache in memory large data structures loaded from DB. 
On client update transactions we perform complex calculations updating them.
Other clients that just browse data, obtain it directly form the memory cache.
In short the same old problem :-(
In the "old" C++ server, this goal was achieved through Shared Memory with an 
impressive speed...

Entity beans would be perfect if I can take advantage from cache bu,t in a 
cluster, right now JBoss doesn't support it.

Is there another tricky way to mantain a "static" global data structure in 
memory except entity beans ? Something that would work on a cluster too.
On a single JBoss instance I could write an MBean but I need a cluster, so this 
data structure should be automatically synchronized through the nodes.

Any suggestion would be greatly appreciate.

Thank You

Davide De Benedictis





-------------------------------------------------------
This SF.net email is sponsored by: Etnus, makers of TotalView, The debugger 
for complex code. Debugging C/C++ programs can leave you feeling lost and 
disoriented. TotalView can help you find your way. Available on major UNIX 
and Linux platforms. Try it free. www.etnus.com
_______________________________________________
JBoss-user mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-user