Databases Reference
In-Depth Information
is delivered via metadata. Metadata identifies an ETL application job and
the data created by that ETL application job. A Metadata application can
then inform downstream ETL applications that their predecessors and
thresholds have, or have not, been met.
Mistakes happen. When they happen you may need to revise rows of a
Fact table. When you need to revise a set of rows of a Fact table you will
need a method that will identify that set of Fact rows separately from all
other rows of the same Fact table. Again, that level of control is delivered
via Metadata. Metadata allows you to identify a specific set of rows of a
table that were created by a specific ETL application executed by a specific
ETL job. Once the errant rows in a Fact table have been identified, they
can be re-created and validated in a separate table. The same dimension
lookup function that yielded an incorrect Instance Key previously can be
repeated. The corrected and validated rows can be repeated using a cor-
rected Dimension table, yielding corrected results. Once the Fact rows
have been re-created and validated in a separate table, they can replace the
errant rows in the Fact table.
The textbook definition of Metadata (data about data) is cryptic and unin-
formative. A better definition of Metadata should include the control, man-
agement, and manipulation of data, ETL processes, BI processes, and data
quality. These functions, features, and methods are explained in Chapter 9
of
Building and Maintaining a Data Warehouse
. Time variant ETL does not
require Metadata. However, the support of a time variant data warehouse is
almost impossible without a Metadata application. So, while Metadata does
not inherently create time variance, Metadata enables a data warehouse to
control and manage the data and processes in a data warehouse.
Instance keys—the manufacturing example
Before we can incorporate sample data from the Iron Ore manufacturing
process first presented in Table 13.10, we need to resolve all the sample
Instance Keys to either Simple Instance Keys or Compound Instance
Keys. Table 13.11 presents all the Instance Keys as Simple Instance Keys.
The Instance Keys in Table 13.11 will be used in the following examples of
the Fact table Manufacture_Metals_Detail.
Table 13.12 includes the Fact table rows found in Table 13.10, with the
addition of Instance Keys for the entities Factor Worker, Raw Material,
Location, and Formula. The Instance Keys, which are Simple Instance
Keys in this example, come from Table 13.11.
