The performance of object reassembly, notably the most expensive operation andthereforemainlydeterminingtheoverallperformance, is clearly crucial for the success of object-oriented databases. In this paper, we examine alternative processing strategies for object reasĆ sembly on a nested relational storage manager (DASDBS) and a flat record storage system (Oracle), as two possible storage subsystems for object database systems. As an example ODBMS we will use COĆ COON throughout this paper. We show how to exploit the complex record features of DASDBS, which allow innovative processing techĆ niques that can result in performance improvements of up to an order of magnitude. In particular, DASDBS as a research vehicle can even outperform industrial strength, well-tuned strategies performed by Oracle.
Christian Rich, Marc H. Scholl