Unfortunately, there is at. present nothing to assist the system architect
at design-time to determine whether a proposed architecture based on an obj
ect-oriented database system will perform as required. The problem is compl
ex, the choice of suitable modelling approach difficult, and a construction
of a model is often abstruse. In this paper we concentrate on a major mode
l component: that describing the access of objects in a centralised databas
e. We present the background for the research (modern corporate IS architec
tures), the choices we have made, the prototype design, and the mathematica
l model of the cost of object database access. We conclude the paper by des
cribing a validation of the model and how it can be generalised. The paper
has a number of objectives: first, to dispel the myth that performance mode
lling of object-oriented systems is an immensely difficult task; second, to
show that techniques which have been in existence for some time for modell
ing are applicable, with some modification, to aspects of object-oriented d
atabase performance prediction; and, third, to detail a specific case study
of access cost modelling which provides enough information to be replicate
d by other workers across a number of object-oriented database products. (C
) 2000 Elsevier Science B.V. All rights reserved.