Conclusion - Data Warehouse Systems: Design and Implementation - page 566

Chapter 15
Conclusion
In this topic, we have provided an in-depth coverage of the most relevant
topics in data warehouse design and implementation. Even though in
Chaps. 11 - 14 we covered advanced and very recent developments, there are
many other important ones that have been consciously left out for space
reasons in favor of mature technologies. We conclude this topic with some
brief comments on these topics, which we believe will become increasingly
relevant in the near future. We refer to a recent topic [ 144 ] where further
perspectives on business intelligence can be found.
15.1 Temporal Data Warehouses
Inmon's classic definition of data warehouses, presented in Sect. 1.1 ,mentions
their nonvolatile and time-varying characteristics. However, in traditional
data warehouses, these features apply only to measures, not to dimensions.
Indeed, although data warehouses include a time dimension that is used for
aggregation (using the roll-up operation) or for filtering (using the slice and
dice operations), the time dimension cannot be used to keep track of changes
in other dimensions, for example, when a product changes its category. This
situation leaves to the application layer the responsibility of representing
changes in dimensions. Temporal data warehouses aim at solving this
problem by extending the data definition and manipulation languages with
temporal semantics. In a temporal data warehouse, changes may occur at
the instance level (as in the example of the product changing its category
mentioned above) or at a schema level , for example, when a dimension level is
added or deleted. Moreover, when the bottom level of a dimension is added or
deleted, the associated fact table is affected, and its schema must be modified.
All of these changes must be automatically handled by the data definition
language. The semantics of the query language must also account for these
changes to produce the correct aggregations.