I agree with Malcolm. You also need to understand the amount of users that will access your application, especially the peak number of concurrent users of the system and tweak it accordingly.
With JMeter you can simulate the number of users you expect to access your app, and see how it performs under load. Keep on increasing the number of users, until the machine cannot handle the load, then start increasing memory in your server and DB. A profiler will also give you stats on where the bottlenecks are. regards Bob On 2011/09/15 01:26 AM, Malcolm Edgar wrote: > Hi Hans, > > The best way to isolate performance issues is to run load tests on a > JVM which you are profiling. It is always surprising where the > performance issues lie. > > I would recommend setting up a load tool (JMeter or something similar) > which has scripts to hit a number of representative pages. Then run > the application in a profiler, or attach a profiler to the JVM and get > some metrics on where the application is spending most of its time. > > With regard to Cayenne, using caching is very important to > significantly improve performance. However caching is a tricky > subject, especially if you are running multiple web applications on a > JVM sharing the same database or across a cluster. This is an area > where the Cayenne documentation is quite poor, so you may need to get > onto the mail list to get some decent advice. > > Also don't use OGNL in your code if you can avoid it. Its performance > is quite poor under load, even when using it intelligently. If you > need to do expression stuff please use MVEL. > > regards Malcolm Edgar > > On Wed, Sep 14, 2011 at 11:49 PM, Hans C. Poo <[email protected]> wrote: >> Hi, >> >> I'm running a click 2.2 application in a dedicated machine with 32GB RAM and >> 4 dual core processors, running in Tomcat 6. >> I would like to hear some concrete advices to exploit at maximum the >> machine, with respect to: >> >> - Cayenne object cache (default 10000) >> - JVM parameters, Xmx, Xms, Xmn, etc. >> - Cayenne cache management >> - Other parameters >> >> We have a separate database server machine with postgreSQL, already tunned >> with pgtune. >> >> The mode in click.xml is in production of course. >> >> I know that this can generate a lot of debate, but i just like to know what >> may be feasible for this installation. >> >> Thanks >> Hans >>
