I've read a lot of references to the performance penalty of using EJBs with remote rather than local references and wanted to get a handle on it. However, my test results surprised me because (JBoss 3.0) it seems to show very little difference in the cost of local vs remote within the same jvm. Of course using remote reference in a standalone client and thus in another jvm was two orders of magnitude slower.
Do conatiners optimize these method call in some way. Is it just JBoss or do they all do it?
For example the create method in milliseconds on Pent II:
local : 0.0841
remote : 0.1051
standalone : 10.706 (101 X greater than remote)