Data Warehouse is a repository of integrated information, available for
queries and analysis (such as data mining). The goal of a data warehouse
is to provide integrated access to multiple sources. The warehousing
approach is appropriate for situations where high query performance
is needed. Therefore, instead of a method capable of accepting
a query, generating the appropriate subqueries for each information source
and merging the results, we are interested in extrating and merging information
from sources in advance and in storing it in a repository over wich
queries are evaluated.
can think of a data warehouse as defining and storing integrated
materialized views over the data from multiple, autonomous and (sometimes)
heterogeneous sources. However, existing materialized view maintainance
algorithms fail in a warehousing environment due to the ``independence''
between the data sources and the data warehouse. In fact, in a warehousing
environment sources can inform the warehouse when an update occurs but
they cannot determine how to do the corresponding update in the warehouse.
Contrary to that, when the context under consideration is just a database
and its views, the source knows what data is needed for updating the views.
Therefore, the propagation of updates of sources to the views at the warehouse
is one important research area in warehousing context.
data warehouses may maintain historical information while the underlying
sources may not maintain this information. Indeed, to make the right
choice for a organization, it is sometimes very important to be able to
research the past and identify relevant trends.
this research we consider the problem of updating a temporal data
warehouse, more precisely the problem of propagating an update over a source
to a warehouse responsible for keeping historical information.