I just read an article from JavaONE which introduced an interesting concept for high-performance caching of read-mostly Entity Beans:
If you read page 20 ff. you'll find the authors proposing the use of two Bean classes: one that is read-only, and another one that is used for write operations. The authors claim performance improvements that seem to be high enough to make this pattern interesting.
Now, I live in Europe and could not visit JavaOne. Can anyone explain the pattern to me? Especially how the surrogate key relates to the production key and where it is created.