Online analytical processing


 

Theory

In theory, Online Analytical Processing (OLAP) stands for hypothesis-supported analysis methods. The analyst must know which requests he/she would like to make on the OLAP system before the actual inquiry. His/her hypothesis will be either confirmed or rejected after the results of the analysis.


Conducting complex analyses, which produce very large amounts of data, is the primary function for OLAP systems. The goal is to produce a result of an analysis, as a support in decision-making, through multidimentional observation of these data.

OLAP systems usually retrieve their data from a data warehouse. This prevents analysis data from coming into contact with transaction-oriented data sets, (if they do it lowers their performance).

Two variations are, firstly, ROLAP ("relational OLAP"), which relies on a relational database and, secondly, MOLAP ("multidimensional OLAP"), which relies on a multidimensional database.

 

Each type has its advantages and disadvantages:

  • MOLAP continually saves the aggregated indicators. Thereby, MOLAP has a performance advantage in comparison to OLAP systems, which aggregate indicators while the calculation is running.
  • ROLAP scales better but the performance of its relational sources is slower than MOLAP.

Practice

In practice, OLAP is mostly used with multidimensional data storage and analysis. Multidimensional databases are therefore also called OLAP databases.

 

Examples of OLAP databases are:

  • Applix TM1 (now Cognos TM1)
  • Infor PM OLAP / Alea
  • Hyperion Essbase
  • LucaNet.Financial OLAP Server
  • Microsoft SQL Server Analysis Services
  • Palo OLAP Server
  • PARIS Technologies PowerOLAP


In contrast to RDBMS, they store data not in tables, but in dimensions, data cubes and cells.

  • Every dimension consists of dimension elements, which are usually arranged hierarchically.
  • Multidimentional data cubes are formed from two or more dimensions.
  • The actual data are then saved in the cells of the data cube.

 

What this multidimensionality means is that multiple dimensions can be applied to viewing and valuating relevant operational indicators (for example revenue or costs); these multiple dimensions may, for example, be customers, marketing channels, regions, and time; the dimensions can also be viewed and valuated on various aggregation levels (for example, segment, product group, product).


XML for Analysis has established itself as the industry standard for access to OLAP databases. It offers data exchange spanning across all manufacturers.


A classic application area for OLAP databases is, for example, multidimensional marketing and distribution planning, or isolated P&L planning with comparatively simple calculation logic. For application in the area of Financial Intelligence, classic multidimensional databases are only partially applicable. Simple relationships may be represented in OLAP databases with the help of so-called rules (for example sales = number x price). This has its limits, however, meaning that complex economic relationships in accounting (especially the interplay of P&L, balance sheet, cash flow statement, analyses and other notes, as well as the context of consolidation according to commercial law) cannot be represented due to technical complications.


As a result, it is not possible to use Realtime Financial OLAP and the especially interactive and efficient workflow on the basis of classical OLAP databases. LucaNet.Financial OLAP Server is an exception, which has been especially developed for use in the area of Financial Intelligence and has a ready operational data model which includes real-time calculation logic.

 

Back to glossary