I have an application which runs in JBoss; I interface with this application over HTTP and send it data for document translation.
The problem is, every minute (give or take a few seconds), I get a spike of .4-1.0 seconds in my transaction log (this is half or double of my normal processing speed; this is a huge penalty). Through lots of research, we've pinned it to an exact point in the JBoss process (well, an exact point in a tusc -p log). You can see the tusc output at http://www.cs.uml.edu/~astanley/tusc.txt and the JBoss thread dump at http://www.cs.uml.edu/~astanley/jboss-thread.txt .
The application in question is Sterling Commerce Gentran Integration Suite 2.1 running on JBoss 2.4.4; the OS is HP/UX 11i running on an RP8400 4 way npar with 4 750MHz CPU's; we have 8GB of ram, 20GB of swap and well in excess of 3T of disk space. JBoss has been moderately tuned, but mostly in the DBconnection pooling department.
Any help would be greatly appreciated!