We continue our discussion on EJB in this post. We would like to talk about persistence.
Persistence can be via Entity beans. Persistence can also be via Hibernate and ORM tools.  We will discuss both. Bean persistence can be managed via two ways: Container managed persistence (CMP) and bean managed persistence (BMP). In the CMP, the container manages the persistence of the bean. In the latter, the developer implements the lower level persistence mechanism. Typically the latter is used when the CMP limits have been exceeded.
In CMP, entity bean data is managed automatically by the container with a mechanism of its choosing. A container implemented on top of a RDBMS may manage persistence by storing each bean's data as a row in a table. Serialization may also be builtin.
A persistence manager is used to separate the management of bean relationships from the container. The container manages security, transactions etc while the persistence manager manages different databases via different containers. Using this architecture allows entity beans to become more portable across EJB vendors.
Entity beans are mapped to a database using a bean-persistence manager contract called the abstract persistence schema. The persistence manager implements and executes find method based on EJB QL.
The persistence manager generates a mapping of the CMP objects to a persistent data store object.  Persistence data stores can vary from relational databases, flat files, and Enterprise Resource Planning (ERP) The mapping is based on the information provided in the deployment descriptor and the bean's abstract persistence schema.
The CMP entity bean and the persistence manager use a contract to define bean to bean, bean to dependent and even dependent to dependent object relationships within an entity bean. When EJB is deployed, the persistence manager is used to generate an instantiated implementation of the EJB class and its dependents using the XML deployment descriptor and the bean class. The instantiated implementation will include the data access code that will read and write the state of the EJB to the database at runtime. The persistence manager also manages the state. The persistence manager generates subclasses for use with the container instead of the abstract classes defined by the bean provider.
CMP is great for database independence and container specific features. The cons are container supported algorithms, portability to other EJB containers, the developer has no access to the view and the generated SQL is not the most efficient.
 BMP pros are that it is container independent, standards based (using EJB and JDBC APIs) , has ability to support nonstandard data types, has flexibility for validation logic, and can take advantage of non-standard SQL features. The cons include the following: it is database specific, requires knowledge of SQL, and takes longer to develop.