A one-to-one association between data and control channels has traditionally existed in transport networks. Being the control plane embedded in the data plane, the design of the former, as well as its resilience, has been addressed in the latter's one. However, a main GMPLS architectural requirement is to provide a clean separation between control and data planes. In this sense, the control plane in GMPLS networks may describe a different topology than the data plane, even realized over a separated IP network. As a consequence of this, data and control network design become no more linked in such scenarios. To the best of our knowledge, no works in the literature have addressed an independent design of the control plane in GMPLS-enabled networks regardless of the data plane. In this paper, we provide a method to obtain the optimal GMPLS control plane design, minimizing the network Capital Expenditures (CAPEX) while matching specific resilience requirements. To this goal, the problem of finding an optimal control plane topology that ensures a certain resilience level is formulated as a non-linear combinatorial model. This model, however, does not scale properly for large backbone networks. In view of this, a constructive linear method is also presented and its optimality validated through simulations on several reference network scenarios. Furthermore, its benefits in terms of total execution time are also highlighted.