Performance issues with a Richfaces application

[sathis.g]

We are trying to isolate performance issues with a Richfaces/JSF application.

On using Rational Performance Tester tool we observed that for every new page request 2 *.jsf artifacts are downloaded (ajaxscripts.jsf, prototypescript.jsf). The size of the above 2 is close to 140kb. The content of these is same across page clicks, however the tester tool shows them as fresh download per page. Other static content (css, image, etc.) are cached and not downloaded with every page access.

Example, (related to attached artifacts)

- Page 1 -Login Submit - Downloaded LoginSubmit-AjaxScript.jsf & LoginSubmit-PrototypeScript.jsf

- Page 2 - Flight Search - Downloaded FlightSearch-AjaxScript.jsf & FlightSearch-PrototypeScript.jsf

We believe this could be a potential reason for performance degradation, hence wanted to understand on reasons for this happening and any potential solutions.


Please let me know if you need any clarification on this

[ilya_shaikovsky]

which app server you using?

[sathis.g]

We are running the application Tomcat 6.0.14

[ilya_shaikovsky]

ok.. I've asked because we have such issue as known for webSphere and this was a server trouble.

In your case all seems should be ok.. Inform us also about RF version and browser you've using.

Also please try script loading strategy defined as ALL.

[sathis.g]

Sorry for the late reply..

But which Script Loading Strategy should ideally be used for better performance. According to the docs,the ScriptLoading strategy "ALL" turnsoff javascript compression.

What is the exact difference between the Strategy "ALL" and "Default" (not able to understnad what is given in the docs)

[nbelaevski]

Hello,

ALL turns off runtime javascript compression because script files that ALL uses are already compressed.

Script and style resources are load on per-component base when DEFAULT is used, e.g. only the resources required by components currently present on the page are loaded.

ALL strategy loads aggregate script and style resources that were created by merging all existing component resources; that aggregate files are somewhat huge in size, but increase performance because the number of server requests decreases. Also if you do not have component C included in view A but use it in view B then switching from A to B won't trigger resources loading for C component if you use ALL.

I suggest that you try monitoring tool built into Eclipse/MS Fiddler tool/some network sniffer and check if resources are cached in fact or not. Maybe performance tester do not cache them for some internal reason?