Stardust supports the following data types that allow the integration within a process:
Serializable data types and Entity Bean data types are Java style data types. They both have their pros and cons. Serialization is a quick method to use objects for your process data. Stardust serializes these objects as whole in the Audit Trail. The disadvantages using serialization are clear. You have to care of different class versions and Stardust have to de-serialize all objects of a type to search for an attribute you have for example defined as descriptor. The Entity Bean Data Type stores only the primary key in the Audit Trail. Stardust does not care of how to persist the object. The EJB container (cmp) or rather yourself (bmp) is responsible to persist the object. But the use of Entity Beans means also the use of a heavyweight persistence framework.
All of these approaches suffer when it comes to queries on the process instance data (as for example if used as descriptors). The engine has to de-serialize/load all entities to evaluate them. So it is not recommended (but possible) to create descriptors on these data types. It is recommended to save descriptor values in primitive or structured data so access to these data is quite fast and efficient. Please also note that sorting is not supported on descriptor values on these data types.