How did you go about selecting the NewSize and survivor ratios? With the initial ratio of 2, each survivor space is 1m and the eden 2m. You will fill up the 2m eden in no time (literally in milliseconds). And the 1m survivor space will be mostly empty because only a small fraction of objects survive a collection.
For production, you should set Xmx and Xms to the same size (prevents the JVM form constantly requesting more memory, or freeing up memory, and from deciding if it should do either. I wonder if that has a bearing on the swap utilization you are seeing, not really sure, it could also be the heap sizes you are using). Do not set the heap to 28G like you have it. You will not be happy if you fill it up and the garbage collector collects that heap - it will take forever (1/2 hour, 1 hour, more???). A good starting point for the heap is 1200M, unless you can show that you need more memory, or can get away with less. Based on your system RAM, 1200M is nothing. If you set both Xmx and Xms to 1200M, you should set NewSize to around 300m (NewSize should be 1/3 to 1/4 of the heap size) . Do not set any survivor ratio arguments unless you have analyzed the data from PrintHeapAtGC and determined that you need to fine tune the survivor ratio. The resulting JAVA_OPS are thus: set JAVA_OPTS=%JAVA_OPTS% -Xms1200m -Xmx1200m -XX:NewSize=300m -XX:MaxNewSize=300m -XX:PermSize=256m View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4077545#4077545 Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4077545 _______________________________________________ jboss-user mailing list [email protected] https://lists.jboss.org/mailman/listinfo/jboss-user
