![]() | Only 14 pages are availabe for public view |
Abstract All existing storage models, for nested relations, have restrictions that limit the database performance. In this thesis, a novel storage model for nested relations -the generalized storage model(GSM)- is introduced. This model, releases all restrictions imposed by existing storage models. Thus, it can locate more efficient storage models. The existing scheme tree is generalized and called the expanded scheme tree. It is used by the GSM algorithm to generate different near-optimum storage models. The algorithm is validated for all the generic query costs that dominate the different nested relational algebraic operations. It is validated analytically for some of these query costs. For the other query costs, it is validated experimentally. Storage models are compared analytically and experimentally. It is shown that, existing storage models, may become near-optimum for certain workloads, while being very inefficient for other cases. The novel storage model is the only storage model that can locate a near-optimum storage structure for any database under different workloads. |