Interfaces Between Modeling Systems and Solution Algorithms

Abstract

Until recently, a modeler needed to know many details about solution algorithms, in the following referred to a ‘solvers’, especially about cumbersome input and output formats, and model building was consequently rather skill intensive and expensive. Although computer times often are quoted as a measure of the effort involved in a modeling exercise, the manpower to implement and debug a model is usually much more costly. Better algorithms has for some time not been the key to cheaper modeling, but rather systems that could manage the model building process. As a result of this demand, several so called modeling systems have emerged over the last years, see e.g. Bisschop and Meeraus [3], Fourer [8,9], and Geoffrion [10].