This paper investigates the application of CARMA, a recently developed quantitative process-algebra-based modelling language, to the stochastic modelling of software defined networking (SDN). In SDN, a single controller (or hierarchy of controllers) determines the behaviour of the switches that forward traffic through the network, and it is used in a variety of settings including cloud and data centres. This research is the initial phase of developing a methodology for agile formal modelling of performance and security aspects of SDN, and focusses on the fat-tree network topology. The results demonstrate that the CARMA language and its software tools which include the MultiVeStA statistical model checker provide a good basis for modelling SDN.