As mentioned before,
Model D
assumes that all attributes
adhere to a uniform distribution.
Hence Model D can use
column unique cardinalities
in estimating selectivities
instead of histograms.
Model D
further assumes
attribute independence:
absent specific
(specified)
functional dependencies,
the values of an attribute
are assumed
to be independent
of all of the other attributes.
Appendix C.4,
Catalog Statistics for TPC-D,
gives the
column unique cardinality,
column minimum value
and column maximum value
for the attributes
of the TPC-D relations
used in the
TPC-D queries.
Notice that
minimum and maximum value
for an attribute
is defined only for numeric attributes.
For many of these (numeric) attributes,
Model D assumes
the column unique cardinality
to be the same
as the cardinality
of the entire table.
For example,
the unique cardinality of
the extended price attribute
in the LINEITEM relation
is assumed to be
the unique cardinality
of of the entire LINEITEM table.
Page 2
