Databases Reference
In-Depth Information
the other as, for example, when one set of data is a summariza-
tion of the other.
A straightforward way to conform data is to derive it from the
same source, at the same time. If transformations are applied to
the data as that data is being copied, then semantically like
transformations, e.g. summaries or averages, should be based
on the same mathematical formulas being applied to the same
data. If identical data is subjected to different transformations,
but those transformations are mathematically mappable one
onto the other, then information derived from those two differ-
ent sets of derived data will be in agreement.
The 2000s
In the first decade of the new millennium, several major
developments took place related to the management of temporal
data. They were:
i. On-line analytical processing (OLAP) data cubes;
ii. Slowly changing dimensions (SCDs); and
iii. Real-time data warehousing.
Data cubes are a software structure, manipulated by business
intelligence software, that provides rapid and easy access to very
large collections of dimensional data, based on often terabyte-
sized fact tables. Slowly changing dimensions are a family of
uni-temporal structures which provide limited support for his-
torical data in data mart dimension tables. Real-time data
warehousing is an evolution from monthly, weekly and then
nightly batch updating of warehouses, to real-time transactional
updating.
Data Cubes
The first of these three developments is of minimal interest
because it is pure technology, involving no advance in the
semantics of temporal concepts. We briefly discuss it because
it is a convenient way to focus attention on the “cube explosion
problem”, the fact that even with terabyte-sized fact tables, a
full materialization of sums, counts, and other statistical sum-
marizations of all the instances of all the permutations and com-
binations of all of the hierarchical levels of all of the dimensions
of the typical data mart would dwarf, by many orders of magni-
tude, the amount of data contained in those terabyte-sized fact
tables themselves. Since including uni-temporal or bi-temporal
data in the dimensions of a data mart would increase the
