In this paper we address the problem of building decision support systems that make use of multiple operations research models as database application. The motivation for developing applications in a database environment is that, by doing so, the development effort can be substantially reduced, while, at the same time, the application inherits valuable database features. the paper contains two main contributions. First, we present a set of modeling constructs that should aid developers in structuring such applications and in carrying out the development process. Included in this material is a fairly comprehensive model for handling versions. Second, we discuss certain design alternatives and evaluate performance tradeoffs associated with hem. In addition, to evaluating the differences among competing database designs, we provide evidence that properly designed database applications, show little performance degradation over file based applications.