Multidimensional, Relational Multidimensional and Hybrid Multidimensional Models

Multidimensional Models (MDM), like Relational models (RMDM), use a multidimensional data model to analyze data. Hybrid models (HMDM) are a combination of relational and multidimensional models. 

In MDM, data is stored in a multidimensional semantic model. It has excellent performance and can-do fast calculations. 

In RMDM, the data is stored in the relational database. It can handle large amounts of data and leverage functionalities inherent in the relational database. Query time can be longer if the underlying data size is large. 

The difference between MDM and RMDM is that MDM requires that information first be processed before it is indexed directly into a multidimensional semantic model, whereas RMDM is entered directly into a relational database. 

HMDM uses a hybrid approach in which queries against materialized dimensions or measures are served from pre-aggregated data, while queries involving non-materialized elements are delegated to relational SQL engines. This model attempts to balance performance and scalability by combining multidimensional processing with relational storage.  

However, semantic layers represent a more advanced approach. Instead of relying on data model storage or hybrid architectures, they deliver governed multidimensional analytics directly on cloud data platforms while maintaining flexibility and scale.  

Kyvos is a cloud-native semantic layer designed to provide high-performance multidimensional analytics on data lakes and cloud warehouses. It enables organizations to define business-friendly dimensions and measures while intelligently managing aggregates for optimized performance. Queries against materialized dimensions or measures are served through the Kyvos Query Engines, while queries involving non-materialized elements can be delegated to external SQL engines such as Presto or Hive. This allows organizations to process a semantic model while selectively keeping high-cardinality dimensions or measures non-materialized — balancing performance, flexibility and scalability without relying on proprietary storage.