In service-oriented software product lines, when a change occurs in the business process variability model, designing the domain architecture from scratch imposes costs of re-architecture, high costs of change in the domain-level assets; and high costs of change in many products based on the new domain architecture. In this paper, focusing on the linear evolution scenario in service-oriented product lines, which refers to propagating changes in some assets to some other assets, both in the domain level, we deal with the problem of propagating changes in domain requirements (the business process variability model) to the domain architecture level, in a cost-optimal and consistency-preserving way. We present a method to suggest the optimal change propagation options to reach the aforementioned goals. The method showed promising to provide minimal change costs as well as to fully preserve consistency of the target models if no human intervention exists in the change propagation results.