On 1/30/06, George Sexton <[EMAIL PROTECTED]> wrote: > Database interaction is by far the major issue. In my application, a page > with no db interaction runs in something like 80ms, while a page with DB > interaction takes something like 350-400ms. There's no silver bullet to > tuning tomcat applications, and most of the advice you'll get here won't > make any difference (use this connector, instead of that). So, my advice is:
Hmm, 80 ms per page is extremely slow. We have a high traffic application here with approx. 100 request / per second / per webserver. Our middle execution time is by 10 ms (both db and not db). Requests without calls to backend are served in below 1 millisecond (mostly ajax). I would say that the most important point is the size of the jsp. Large jsps need time, and CPU. Other points are: Avg. Request duration.( see above). With 30 requests per second you can't afford spending more than 33 ms per request. Connection Speed: If you have keep-alive on, the thread will be blocked until the data is delivered. Http 1.1 (keep-alive) connections consume 1.5 threads per connection. My advice: use http 1.0. Zipping can also reduce thread consumption (but increase cpu consumption). WIth all above said, if you can serve fast enough, use no http1.1 and zip, you can use tomcat default settings (max 150 threads). Alas, do you know your read/update ratio for the db? If you have 80-90% reads on the DB, you can win a lot of performance through caches. Regards Leon --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
