Evaluation of storage mechanisms for a content repository

In this thesis, we describe and evaluate different storage mechanisms for the content repository powering a content platform. In certain usage scenarios the current RDBMS-based implementation exhibits severe performance degradation. The evaluation will focus on performance and implementation complexity. To perform a realistic evaluation, we have gathered real-world usage scenarios and data statistics. In an ex- periment, three prototypes were created and benchmarked based on these usage scenarios. The experiment results show that the object-oriented database db4o does not scale well when used as a back-end for a content repository. Additionally, the graph-oriented HypergraphDB database was shown to scale to large datasets but does not provide ordering and pagination, two important features for a content repository. The results suggest that a document-oriented database is a good alternative to an RDBMS as the back-end of a content repository in terms of both performance and implementation complexity. A disadvantage of this alternative is the reduced durability.