Hi Randy, Thank you for your reply.
> > One common misconception that people seem to have about Tomcat (and > application servers in general) is that you can predict how well an > application will perform on that server without any real details of the > application, but exact hardware details. > I understand that it is a vague and open ended question and I mentioned it so in the email. What I am looking for is a ballpark figure that might not work for me but will give me something to shoot for/surpass. Not having worked on Tomcat or any other app server performance issues leaves me asking this question. > For almost all applications that Tomcat is going to serve, the > majority of CPU time will be spent in your application. The majority of > network traffic will be specific to your application. The majority of > memory might be your application (Tomcat might actually tie with you here). > The point being, of all the possible factors that can limit the transactions > per second or whatever other performance metric you want to consider, your > application is the biggest blocker, not Tomcat, not the OS. > That is true. My application does require network traffic. But I have used network monitoring tools like ntop etc. and am sure that network is not a bottleneck for this application. About the memory, I have used OptimizeIt and know where the memory bottlenecks are and am confident that garbage collection takes care of it (though there are some places that can be tuned) > My best advice is for you to try it out. Set up a load test and > see. One potential bottleneck you didn't mention, by the way, is your > bandwidth between you and your users - this can also limit the number of > transactions (which is a rather vague term when discussing app servers) you > will receive. > I have used load tests and for a single test it takes close to 3 min and have tested for 500 instances of the same test at the same time achieving decent performance. Bandwidth is not bottleneck bcoz all users are on T1 line. So what I am looking for are ballpark figures. Hope this information helps you and other tomcat list users to estimate the application. For the benefit of other "listers" I am pasting the email I wrote earlier below. ************* > > > > I have a single servlet based application. This application > > is accessed > > througha browser and travels over to a high performance > > database through a > > RMI-IIOP connection. It is not the best application in terms > > of code tuning. > > We are deploying this application on Tomcat 4.0.1, Apache > > 1.3, Suse Linux > > 7.0 on a PIII 800Mz machine with 800Mz RAM. We are not > > planning on adding > > multiple instances of Tomcat/Apache for the time being (or we > > are not quite > > sure that our machine might be able to sustain that many > > threads of either) > > > > Based on this what is the maximum number or requests that I > > can expect to > > obtain from my applications. Can I expect say 1000 (or maybe > > 5000 or more) > > concurrent users. How many requests/sec can I expect while delivering > > acceptible response on the browser. > > > > The reason I have asked such an open-ended question is that > > you might have > > deployed your production system on such a configuration and must have > > achieved some bechmarks. Please take the time and give me an estimate. > > > > Thanks in advance. ***************** Thanks Brown. -- To unsubscribe: <mailto:[EMAIL PROTECTED]> For additional commands: <mailto:[EMAIL PROTECTED]> Troubles with the list: <mailto:[EMAIL PROTECTED]>
