tomcat-taglibs-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Henri Yandell <>
Subject Re: JSTL startup times - is caching going on?
Date Sun, 21 Sep 2003 05:45:14 GMT

The JSP spec for Tomcat 4.1 does say that a container may cache Tag
objects between page runs. It sounds like this could be what's happening
here. I've no idea whether it would cache across sessions etc. Up to the
server and not the taglib.

On the speed issue. What are you using to create the SVG? And where does
it get the data?

Off-topic: I've been having fun with the Cewolf JFreeChart library and
it's pretty nifty.


On Sat, 20 Sep 2003, Richard Mixon (qwest) wrote:

> I am having an issue with large startup times the first time one of my
> pages runs that uses JSTL mostly the core library. Subsequent requests
> (even for different users on different data) take much, much less time.
> Here is a little more detail. Any ideas are much appreciated. My ISP is
> pretty irate because we suck up huge amounts of CPU each time it
> happens.
> We are using Tomcat 4.1.24, Java 1.4.1 and Struts 1.1.
> OK, the page uses a couple of nested foreach loops to generate a
> grid/chart of SVG markup language. It also uses a couple of small
> foreach loops to generate tick marks on an axis. It looks something
> like:
>   Chart  page
>     <embed> tag
>     include JSP/JSTL page to generate footer.
>     include JSP/JSTL page to generate x and y axis
>     JSTL markup to generate main body of chart
>     include JSP/JSTL page to generate footer.
> The request takes a half second to build the Java objects that the JSTL
> code uses to render on the page. But the rendering takes 45 second on a
> 650MHZ Sun SPARC III machine and almost that long on a fast Windows
> machine (2500 MHZ).
> The second and subsequent times the chart rendering drops down to less
> than 200 milliseconds. Then about five minutes later, if the chart is
> requested again, it takes the 45 seconds to generate, then 200
> milliseconds for subsequent requests.
> Thanks - Richard

View raw message