Choose appropos Data Access!

Popular Methods for Data Access

•JDBC API –a CLI view of RDBMS data
•Apache iBATIS–simplifying common JDBC usage patterns
•EJB v1, v2 –a "record oriented"approach
•Object Relational Mapping (ORM)
–Hibernate, Castor
–JDO v2 (Including KODO, OpenAccess)
–EJB v3 (Including Hibernate, Toplink, KODO)

Data Access: JDBC

•The "assembly language"of RDBMS –you can do anything, but you have to do everything
•Best choice when the form of the data being accessed is unknown, such as in a reporting engine in which the number and types of result columns are unknown
•Good choice for dealing with extremely large result sets and accessing rarely-used driver functionality
•Worst choice for rapid application development
•Worst choice for large engineering teams and large code bases
•Worst choice for building maintainable applications


Data Access: ORM

•The "object oriented"model for RDBMS –you only deal with objects, but the ORM has to deal with the RDBMS
•Best choice when the form of the data being accessed is well known, and the wide-spread use of the data throughout the application logic far exceeds the investment in defining the object schema and its mapping to the database
•Good choice for enabling data caching
•Definitely nota silver bullet
–Still requires good development processes and careful design
–Using ORM, some common application use cases are very inefficient compared to hand-coded JDBC


Why the Choice is so Critical

•I have witnessed more applications fail to meet their business goals due to poor choices around data access than any other category
•Once a choice is made, it tends to be reflected in every aspect of the application, making later changes more difficult and incredibly costly
•The predictability and cost of scalability of a large-scale application is tightly bound to the application’s data access model


Choosing a Data Access Model

•Understand the high-level requirements
–The "-ilities": Scalability, Reliability, Availability, …
•Visualize the data flows in the running system
–For each page or service request, what actually goes through to the database, and why?
•Understand the impact of concurrent users
–How will database contention be minimized?
–How will cache effectiveness be maximized?
•Understand the application’s data granularity
–Set-centric or identity-centric?