One of the main impediments to large OpenFlow deployments is the insufficient number of flows in today's hardware switches. That being said, the number of flows is not the only stat that matters in a network implementation. Ivan Pepelnjak takes a more nuanced look at how newer versions of OpenFlow (1.1 onward) introduce the concept of tables, which seem a lot like database tables since they're used for recording reference data.
OpenFlow 1.0 is too barebones to be flexible enough for all the future implementations you might have to build. And OpenFlow 1.3 has even more flexibility features as it just begins to come into the market. Pepelnjak's advice: wait and buy an OpenFlow 1.3 switch.