The Data Warehouse Development Life Cycle
Online Analytical Processing and Oracle
Relational OLAP (ROLAP)
Because the entire enterprise can be
made available to a ROLAP tool, it is not surprising to acknowledge
that ROLAP is far more flexible than its MOLAP cousin. Any data, on
any platform or database, can be mapped into a multidimensional
format; the ROLAP engine will obediently extract and summarize the
data according to the extraction specifications. Because ROLAP is
far more robust in this sense, it is the OLAP tool of choice for
data warehouses that support the following features:
*Data changes frequently--In
a data warehouse where data is very dynamic and end users require
up-to-the-minute summarizations, ROLAP is the only choice. MOLAP
tools must extract and summarize data offline for loading into their
multidimensional databases. To make matters worse, most
multidimensional databases require recalculation of the entire
database when a new dimension is added, an aggregation scheme
changes, or data is added. These overhead factors make MOLAP
inappropriate for decision support systems with highly volatile data
sources. Examples of these types of applications include stock
market DSS and weather forecasting tools.
*Large data volumes--For
very large database warehouses in the terabyte range, the cost of
supporting MOLAP tools can be exorbitant. The pre-summarization of
data can require hundreds of gigabytes of disk storage, and many
companies cannot afford the millions of extra dollars required to
provide sub-second response times for OLAP queries. ROLAP tools
allow companies to leverage their existing investment in OLTP
databases without having to buy a multidimensional engine.
*Unpredictable types of
queries--Because ROLAP engines can allow virtually any
operational data source to be queried and summarized, ROLAP has a
clear advantage for the decision support application that cannot
predefine its query requirements. Of course, this flexibility comes
at the cost of ease-of-use, because the IS department must often get
involved to assist end users in creating the mappings to the
operational databases.
Today, many developers are using
relational databases to build their data warehouses and simulate
multiple dimensions, and specific design techniques are being used
for this. The push toward STAR schema design has been somewhat
successful, especially because designers do not have to buy
multidimensional databases or invest in expensive front end tools.
Several methods can be used to
aggregate data within OLAP servers. As you can see in Figure 5.18,
this method extracts data from the relational engine and summarizes
the data for display. Another popular method pre-aggregates the data
and keeps the summarized data ready for retrieval.
Figure 5.18 Aggregation and OLAP servers.
and include the URL for the page.
