As a DBA I have always used stored procedures to access database data for the myriad of reasons that we all are aware of. My current project has made the decision to forgo stored procedures and access data directly from the underlying tables by means of JPA.

I understand the concept of persistence, but I have seen any analysis that addresses my primary concerns of performance, reusable code, security, and standards. Everything that I have learned about database design over many years is that you never allow direct table access and to control who can access the data via stored procedures.

 With the JPA approach, tables are opened up for direct access. From what I have seen adhoc queries are buried in the objects. I have heard but not been able to validate is that the objects create cached query plans to allow for operations similar to that of stored procedures. Are there any studies that formally compare data access performance of stored procedures as compared to JPA?

Also, with these query statements buried in these objects does this not eliminate the concept of reusable code? In the development of stored procedures I typically create a set of low level tools that all procedures can use for common operations to eliminate duplicate code. Unless there is something that I am missing it would seem that this reusable code construct is eliminated and each different query would need to repeat common operations.

Maybe this is the new frontier but I have not seen and can not find any comparative studies that analyzes the performance of JPA to that of stored procedures especially with large complex databases.

Is there anything available to indicate that replacing stored procedures with JPA is the correct direction to take?