The application of functional networks in the automotive industry is still very slowly adopted into their development processes. Reasons for this are manifold. A functional network gets very quickly complex, even for small subsystems. Furthermore, concepts for variability handling are not sufficient for future standards such as AUTOSAR. In this paper we will provide an approach to reduce the size of funcetworks by applying abstraction and partitioning. The achieved improvements will be described. In addition we will provide an alternative concept to handle variants in functional networks. The approach is based on extracting variation points from the functional network and modeling them separately. Open problems and future work are discussed at the end.