Multidimensional OLAP (MOLAP)

Multidimensional OLAP is generally thought of as the traditional multidimensional database (MDDB), and many of the early offerings advertised themselves as “pure” multidimensional databases. As we have discussed, a multidimensional database is a database structure optimized for storing facts categorized along many dimensions. The MDDBs are far more effective for storing OLAP data than relational databases because they were designed exclusively with this purpose in mind. The other major consideration with multidimensional OLAP is the fact that all of the data is loaded, summarized, and stored in the MDDB prior to making the database available to end users. Because all the calculations have already been performed, multidimensional OLAP offers astounding response times. For these reasons, multidimensional OLAP is the best choice for applications with the following characteristics:

*Impatient end users--MOLAP engines offer end users fast and predictable response times for their queries. In some cases, end users need to be able to quickly create new queries based on the responses from previous queries without losing their train-of-thought. This speed differential is getting smaller as the speed of relational databases improves, but there remains a dramatic difference between the retrieval of pre-summarized data from an MDDB and the runtime extraction and summarization from a relational back end. It is not uncommon for an end user to report a system outage when the ROLAP tool takes several hours to roll-up summaries from a relational database extract.

*Sophisticated data analysis--MOLAP engines provide a more robust analysis environment than ROLAP tools. MOLAP engines support budgeting and forecasting functions and tend to have a much more advanced statistical toolkit than their ROLAP cousins. ROLAP, on the other hand, has the ability to provide ad hoc groupings while MOLAP cannot aggregate on the fly.

*Ease of use--MOLAP engines are very easy for end users to configure and use to set up scenarios for decision support systems. Because the data is pre-summarized and stored in the multidimensional database, all an end user needs to do is specify the dimensions and groupings within dimensions. ROLAP, on the other hand, requires an end user with knowledge of the mapping of the operational databases, and it is much more difficult to configure.