Metrics for Dynamics: How to Improve the Behaviour of an Object Information System

Abstract

If we ask about which is the main difference between modelling a system using a traditional model like the entity relationship model or an object oriented model, from our point of view the answer is that, in the first one, the processes are not located somewhere, and, in the second one, the processes (operations or methods) are encapsulated in classes. The choice of the right classes to home every operation is essential for the behaviour of the system. It is totally useless to design a well built system, according to a lot of statics metrics, if the system does not run well after. In other words, dynamic metrics allowing to evaluate the behaviour of a system when it runs are much more useful than any static metrics used to tell if the system is correctly built or not. According to this, we propose in this paper, a new approach to evaluate a priori the behaviour of a system, by taking into account the notion of event cost and the notion of time (which is obviously essential). The final goal of this approach is to deliver information on the way operations have to be placed in classes in order to get better performances when the system is running. However, the proposal of metrics is of no value if their practical use is not demonstrated, either by means of case studies taken from real projects or by controlled experiments. For this reason, an optimisation tool is being under construction in order to provide solutions to this problem