Optimization methods for reliability systems on diversified criteria for varied types of models are being increasingly developed. Venugopal et at. (1986–1989) have initiated notable lines of work in this direction for a large class of reliability models considering some comprehensive types of cost functions. This paper adds to this work in terms of considering the complex models, namely the series–parallel reliability system. A methodology is developed with the help of an algorithm to obtain cost-space optimal systems in a class of series parallel models based on practical cost considerations. The operational use of the theoretical work is brought out through some empirical results. A qualitative analysis is also provided as a useful companion study.