Fig. 12.1.2 The essence of operational processing.Operational systems have the mission of being speedy execution, operation against data
at a detailed level, and bound together in applications.
Because of the need for speedy execution of transaction, data are often denormalized.
Denormalization is the design technique that the designer needs to use in order to
enhance performance. But because data are denormalized, they are “pulled apart.” One
unit of data is found in one database, and the same unit of data is found in another
database. The fragmentation of data into separate databases is a natural result of the need
for denormalization of data in a high-performance environment. Denormalization of data
in the high-performance transaction processing environment is a normal, natural
phenomenon.
But there is a side effect of denormalization of data. Because data are denormalized in
the operational environment, data are not integrated. The same unit of data often exists in
several places. (Or in the worst case, the same unit of data exists in many, many, many
places.) The net effect of the same data existing in many places is that the data lose their
integrity. One user accesses the data in one place and gets one value. Another user
accesses the same data in another place and gets a very different value.
Both users think they have the correct value of data. And both users have very different
values.
Chapter 12.1: Operational Analytics