Dimensions of Integration

Integrated databases can be roughly classified as points in a plane defined by two axes:

• Loosely to tightly federated
• Materialized to virtual view

Loosely vs tightly federated:

• How many tasks of integration have been performed, and in how much detail
• A tightly federated database is one in which
  • the schemas have been transformed into a common model,
  • as much semantic schema matching and merging has been performed as possible, and
  • data transformation and semantic data matching has also been performed.
• Advantages of tightly federated database are:
  • Single data representation uniting everything
  • Single query language for developer and end-user
  • Reduced ambiguity
• Disadvantages of tightly federated databases are:
  • Costly to build
  • Costly to maintain
  • Inflexible

Materialized vs virtual view:

• Whether there exists a physical copy of the integrated databae, or whether the integration describes how to translates queries against a view into queries against the underlying data sources
• Advantages of materialization:
  • Faster
  • More reliable
  • Easier to enforce inter-database constraints
• Disadvantages of materialization:
  • High initial cost
  • High maintenance cost

Note that the same methodology/tool can be used to implement both ``tightly integrated and materialized'' databases and ``loosely integrated virtual'' databases.

The more flexible a methodology/tool is, the more useful it will be since it will admit a gretaer range of solutions in data source integration.