It’s all run by a controller that’s centralized but also includes a federated piece distributed to each switch. The setup is similar to the way OpenFlow gets deployed, but the inner workings are very different (and no, OpenFlow itself isn’t supported yet). Plexxi uses algorithms and a global view of the network to decide how to configure the network.
In other words, rather than programming route tables, the controller looks at the needs of the workloads and calculates how the network ought to be getting used. Some of this can even happen automatically.