> the question is, how are you constructing the HTML? Are you opening a
> template file locally, and then using string replace? Or are you using
> XSL/XSLT with a stylesheet to do a stream parse/transform?
Neither.
Since I'm working with a simple login page, most of my text is strings,
appended to a stringbuffer to create the login page.
That login page is then saved as a global, and retrieved for all subsequent
requests.
The information that is unique in every user's request (one string)
is modified with a StringBuffer.replace in a tag that is recorded as part of the
global login page.
output is done using the following psuedo code
is page cached
StringBuffer = cached page
replace tag
println(stringbuffer)
println(end of html page) (</FORM></HTML)
> What kind of system was the benchmark performed on. The page should
> only take 7-10ms properly written. If you're never going to get more
> than 10 concurrent requests, you're probably not going to reach the
> point where threads provide benefit. Especially if it's pages that do
> not use databases or perform other processes which benefit from
> threading.
ab was on a freebsd box, servlet was on a redhat 6.2. As far as concurrent
requests, I've got to be able to handle close to 100 requests, but as I said
earlier I don't think threading is the issue as even single requests show very
poor performance relative to a CGI.
Aryeh
---
Aryeh Katz
VASCO
www.vasco.com
--
To unsubscribe, e-mail: <mailto:tomcat-user-unsubscribe@;jakarta.apache.org>
For additional commands, e-mail: <mailto:tomcat-user-help@;jakarta.apache.org>
