Star/Snow-Flake Schema Driven Object-Relational Data Warehouse Design and Query Processing Strategies

Abstract

The conventional star schema model of Data Warehouse (DW) has its limitations due to the nature of the relational data model. Firstly, this model cannot represent the semantics and operations of multi-dimensional data adequately. Due to the hidden semantics, it is difficult to efficiently address the problems of view design. Secondly, as we move up to higher levels of summary data (multiple complex aggregations), SQL queries do not portray the intuition needed to facilitate buildingand supporting efficient execution of complex queries on complex data. In light of these issues, we propose the Object-Relational View (ORV) design for DWs. Using Object-Oriented (O-O) methodology, we can explicitly represent the semantics and reuse view (class) definitions based on the ISA hierarchy and the class composition hierarchies, thereby resulting in a more efficient view mechanism. Part of the design involves providing a translation mechanism from the star/snowflake schema to an O-O representation. This is done by flattening the fact-dimension schema and converting it to a class-composition hierarchy in an O-O framework. Vertically partitioning this O-O schema further increases the efficiency of query execution by reducing disk access. We then build a Structural Join Index Hierarchy (SJIH) on this partitioned schema to facilitate complex object retrieval and avoid using a sequence of expensive pointer chasing (or join) operations.

This work is supported, in part, by The UGC Research Grants Council of HKSAR under grant CityU 733/96E.