Skip to main content

Total througput is a sum of individual web application throughputs?

1 reply [Last post]
Joined: 2007-04-03

Java Performance Specialists,

We are in the process of performance testing, analyzing and tuning a Java/JSP based web application running on an application server. The application server has 8 logical instances running to serve the user load.

Farm of 8 application servers in total is suppose to provide a total throughput of about 40 Pages /Second. ( calculated top down from capacity planning calculations)

We will deploy two large web applications on this farm each serving a different set of web community and both are being performance tested and tuned individually.

We are trying to find a best way to test and tune individual web sites while using correct performance metrics to achieve, the question we have is -

While we are testing and tuning both the above web applications separately, should we make each one of them comply to a throughput requirement of 40 pages /sec or should we arrive at a proportional figure for each separate application, which can be say 15 pages / sec for first web app and 25 pages / sec for the second web app, and test and tune the web site individually for these metrics?

The latter seems to make sense to some of us as throughput is a metric with upper bound. If load on one web app is not enough to reach the collective throughput, it would be unreasonable to keep increasing the load on one web app ( beyond it's peak load) to reach the required throughput. We should rather apply peak loads on both the web app separately and make sure collective throughput is 40 pages /sec. But we want to make sure we are thinking on right lines.

The way we will deploy the appications is that mostly the first one would be deployed on 4 managed server and the second one on other 4 managed server on Sun blades sepparately.

Would appreciate an early reply with some reference if possible.

Thanks for your help!

Reply viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
Joined: 2005-05-23

The real answer depends on who is setting your requirements. Where did the 40 pages/second come from? Did it originally come from adding up the pages/second of the individual apps? Is it just some random number plucked out of the air, or is there some rationale behind it?

To me, measuring the individual apps is the only thing that makes sense, and each app should have its own throughput requirement. If users are expecting 15 pages/sec for app A, it doesn't help them to know that app B is getting 39 pages/sec and thus you're meeting 40 pages/sec because app A is getting 1 page/sec -- they're only getting 6% of the throughput they are expecting. So I'd completely ignore the aggregate pages/second (except I suspect that requirement came to you from someone who isn't thinking clearly but still has power over you :-) -- still, if you meet the individual requirements, you'll have met the overall (somewhat questionable) goal while doing the right thing by your users.