Joe Ess wrote:I've used The Grinder to measure throughput and response time, which are probably the metrics you are interested in.
Our Performance FAQ has some starting points for working on performance.
I'd add one performance tip to that list: If you want "Enterprise level" performance (i.e. thousands of users, large data sets, complex operations), make sure you are running on "Enterprise level" hardware. I've seen no shortage of offices that retire an old PC that isn't suitable for desktop use to a closet for use as a "server", then come back to me and complain that "java is slow". A $10k investment in hardware will improve performance much faster/easier/cheaper than employing me to tweak SQL queries and loops for a month.
Thanks for your reply! Yes, I understand that hardware can always be a barrier to the performance tuning and I have seen lots of example already. In my case, for now I am not worried about that level as we are using vast configuration actual server. I would like to try to this tool that is mentioned above, and also I found that the article that you shown was very effective and found interesting point (18) in this article.
In our application, there is one such large which runs fine on firefox, but hangs up in IE.
Recently, we upgraded tomcat to tomcat6.0.18 and found the minimum memory used by tomcat is improved a lot in fact. tomcat6x is lots better.