Using object-oriented databases to preserve integrity and improve performance - But at what cost?

Two alternatives to using object-oriented databases to improve performance of complex relational databases are, firstly de-normalisation which improves performance but at a cost of additional errors and/or code to be maintained, and secondly, obtaining a bigger and faster computer though this may not give the same improvement. This paper identifies a need to properly cost the reported performance advantages of object-oriented databases and to compare these costs with alternative performance improvement methods.