| Data remains in data sources and queries are
translated from a common data model to queries that each data
source can execute. Queries are executed on, and results
retrieved from, data source systems. The components that
translate queries and transform result-sets are called adapters. 
Figure 1: Federated database system
diagram.
Advantages
Federated database systems
were pursued in an attempt to overcome some of the disadvantages
of data warehouses by providing the following primary benefit:
 | Data remains at source |
The above benefit overcomes
many of the data warehouse challenges of:
| Complex
ETL process (time and cost) - see comment below* |
| Data ownership issues |
| Static and dated data |
| No drill-down
capabilities |
Disadvantages
However, there are
considerable disadvantages of federated database systems that
generally counter data warehouse benefits due to data source
constraints:
| Dirty data as is |
| Limited indexes not
consistent across data sources and not flexible |
| Limited query processing |
| Query load on data source
system |
| Query performance |
| Security |
| Data source owners aware
of queries (intelligence-related) |
To accommodate the translation
between an application or information sharing system and any
particular data source, conventional adapters are developed,
typically, over a significant period and at great cost to cover
basic requirements. In fact, it typically costs 300 to 500% of
the initial adapter purchase cost to customize conventional
adapters to cover basic requirements.
*The only advantage
conventional adapters have over the ETL process is that schema
transforms are not as difficult; however, query processing and
subsequent results transforms are more complex. |
