Question : Will moving to JBOSS result in the claimed performance improvements?

Consider these claims

http://labs.jboss.com/file-access/default/members/jbossweb/freezone/index.html

Currently we are having a fairly simple app (jsp, servlets, oracle db on an oracle raq cluster, application uses a few simple tables) running on two high end servers fronted by a loadbalancer.  Presently there is no session replication across the servers.  Apache httpd handles the ssl encryption/decryption and uses AJP13 to talk to a tomcat instance on each server.  This config is ok for most of the time except when there are huge surges (maybe 5 times a year) in which case we have to present a slimmed down version of the site.

Will moving to JBOSS  (and dropping  apache for SSL) give us twice the performance as claimed?  What are the potential pitfalls/advantages (gotchas)?

The OS is RHEL  but we do not want to commit to this config long term (if there are special binaries involved).

I have also heard about Terracotta recently and if you have an opinion thats also welcome (although note that presently we dont cluster sessions at all as that would further reduce capacity of the servers)

p.s. In a weird twist of fate, Experts Exchange was restarting its servers as I posted this question.. Luckily I had pasted it into notepad2!

Answer : Will moving to JBOSS result in the claimed performance improvements?

You could either:

   1) Redo all hrefs so they specified http://servername/file
   2) Move all static content to another server and redo hrefs to http://newserver/file
   3) setup re-writes so that https://servername/file gets re-written to http://servername.file

The last requires the least amount of changes, but uses the most CPU because Apache must do re-write processing.  The 1st options means you don't have to setup a new server, but still uses some CPU on your application server for non-application processing.  The 2nd option offloads the processing of static contents and images.

You may want to check your load balancer.  Some can be configured to redirect HTTP requests based on the URL being referenced.  So you might be able to setup a new server, rewrite the URL so that it is http://newserver/file  instead of https://currentserver/file and only need to setup a new server and move the images.

Using distributed boxes it is best to get as much processing as possible off the application server.