Conclusion - Data Warehouse Systems: Design and Implementation - page 567

Temporal data warehouses aim at applying the results of many years of
research in temporal databases to the data warehouse field. Temporal
databases provide structures and mechanisms for representing and managing
information that varies over time. In short, temporal databases allow past
or future data to be stored in a database as well as the time instants when
the changes in these data occurred or will occur. Thus, temporal databases
enable users to know the evolution of information required for solving complex
problems in many application domains in which time is naturally present, for
example, land management, financial, and healthcare applications.
Temporal data warehouses raise many issues, including consistent aggrega-
tion in the presence of time-varying data, temporal queries, storage methods,
and temporal view materialization. Further, very little attention has been
given by the research community to the conceptual and logical modeling
of temporal data warehouses or to the analysis of the temporal support
that should be included in data warehouses. Some of these issues have been
addressed in the literature to various extents.
Golfarelli and Rizzi provide a survey of temporal data warehouses [ 66 ].
With a focus on conceptual modeling of temporal data warehouses, Mali-
nowski and Zimanyi [ 124 , 127 ] introduced time-varying (i.e., temporal) data
types for keeping the history of data warehouse dimensions and extended the
MultiDim model studied in this topic to address temporal data warehouses.
Also, translation rules from the conceptual to the relational and object-
relational models are given. Logical data models have also been proposed
for temporal data warehouses [ 42 - 44 , 136 , 137 ]. For example, Mendelzon and
Vaisman [ 136 , 137 ] introduced TOLAP, a data model and query language
where the schema and the instances of the relationships between levels in a
hierarchy are timestamped with their validity intervals. These timestamps
define how dimension-level members are aggregated. In this way, we can
aggregate measures according to the dimension schema and instances that
existed when the corresponding facts occurred.
It is worth remarking that slowly changing dimensions, studied in Chap. 5 ,
address the problem above in a limited way and are only one variant of
temporal data warehouses. Further, the slowly changing dimensions solutions
do not take into account all the research that has been done in the field
of temporal databases. We have seen that some of the solutions for slowly
changing dimensions do not preserve the entire history of the data and are
dicult to implement. One of the main differences between the temporal
models discussed above and the slowly changing dimensions approach is that
the semantics of the updates to dimension hierarchies is ignored by the latter.
Thus, the valid path at a certain instant in a temporal hierarchy must be
computed manually at the moment of writing the query rather than being
accounted for automatically by the query language.
Another approach to temporal data warehousing is multiversioning. For
example, Ravat et al. [ 170 ] defined a multiversion multidimensional model
that consists of a set of star versions, each one associated with a temporal