| Previous | Table of Contents | Next |
Online Analytical Processing (OLAP) is a class of analytic application software that exposes business data in a multidimensional
format. This multidimensional format usually includes the consolidation of data drawn from multiple and diverse information
sources. Unlike more traditionally structured representations (for example, the tabular format of a relational database),
the multidimensional orientation is a more natural expression of the way business enterprises view their strategic data. For
example, an analyst might use an OLAP application to examine total sales revenue by product and geographic region over time,
or, perhaps, compare sales margins for the same fiscal periods of two consecutive years. The ultimate objective of OLAP is
the efficient construction of analytical models that transform raw business data into strategic business insight.
There are many ways to implement OLAP. Most OLAP systems are constructed using OLAP server tools that enable logical OLAP
structures to be built on top of a variety of physical database systems, such as relational or native multidimensional databases.
The following features are generally found in most OLAP systems:
• Multidimensional representation of business data.
• Upward consolidation of multidimensional data in a hierarchical manner, possibly with the application of specialized processing rules.
• The ability to navigate a hierarchy from a consolidated value to the lower level values forming it.
• Support for time-series analysis; that is, OLAP users are generally concerned with data and consolidations at specific points in time -- By date, week, quarter, etc.
• Support for modeling and scenario analysis -- A user should be able to apply arbitrary “what-if? analyses to a result set without affecting the stored information.
• Consistent response times, regardless of how queries are formulated -- This is critical for effective analysis and modeling.
OLAP applications integrate well into the data warehousing environment, because a data warehouse provides relatively clean
and stable data stores to drive the OLAP application. These data stores are usually maintained in relational tables that can
be read directly by OLAP tools or loaded into OLAP servers. These relational tables are often structured in a manner that
reveals the inherent dimensionality of the data (such as the ubiquitous Star and Snowflake schemas). Also, the data transformation
and mapping services provided by a data warehouse can be used to supply OLAP systems with both metadata and data. Transformation-related
metadata can be used to track the lineage of consolidated OLAP data back to its various sources.