OLAP OF THE FUTURE

On-line analytical processing (OLAP) is a powerful data analysis technology. However, current implementations bypass the 3-layer model of databases for performance reasons and create a redundant database The clear separation between external views, conceptional model and internal physical representations in the three layer database architecture constitutes one of the key factors to the success of database technology for on-line transaction processing (OLTP). However, experience has shown that it is not feasible to construct an on-line analytical processing (OLAP) view onto the data within this model. As a compromise, materialized views were introduced. Yet, these limit the applicability of OLAP to fairly standard applications. Some application domains such as the analysis of stock market data or telecommunication systems with a diverse structure and unpredictable aggregation and query patterns can not be timely handled using pre-materialization. As a result, the benefits of OLAP are not available when building analysis and visualization tool for this kind of applications. We envision OLAP-on-demand as a foundation for the re-unification of the transaction processing and analysis in databases. Instead of pre-materializing expected queries, analysts can integrate the available data sources on the fly where the analysis of information is not delayed by certain ``update windows''. Although current computing hardware does not provide the necessary data throughput, we expect future generation systems to cope with the high demands of this approach. In the meantime, parallel and distributed query processing can be used to achieve the required response times. As an ultimate goal the realisation of Codd’s formulaic model is envisioned