You bring up a interesting point, and one that I struggle with. Im not sure if the network of tomorrow absolutely needs some of this stuff. With the cost of 10 gig and 40 gig optics coming down, a strategy of overbuilding the network could easily take the place of having to dynamically reroute traffic. Large swings in traffic could still be possible even if they weren't predictable. Also, as a network engineer, Im not super keen on letting VMs move wherever they want so we generally keep vmtions isolated within DC pods. I still call layer 2 domains failure domains
The other piece I struggle with is the controller itself. If the reason for having a controller is so that end users can program their own network logic I dont think I'm a fan. Like a guy I work with says, "Do you really think you can do network logic programming better than Cisco and the other big network players?". Can some of us? Im sure Google and Amazon can, but Im not sure I can.
Interesting discussion though. I often think that much of this network revolution won't change the way I do my job. Put another way, Im really only interested in new network tech that fixes the problems I currently have. Im fine with the way my core and distribution switches function on their own now (in most cases). However Cisco ACI is peaking my interest because it fixes many of my problems (Intelligent network fabric that doesnt have the flood mentality).