The second day of the SDN World Forum was again very interesting. My main goal for today was to get a better understanding of Network Functions Virtualisation (NFV). From what I learned today NFV is nothing more than a virtualised systems on commodity (x86) hardware. In that it does not differ a lot from the current cloud based solutions like Openstack. The difference is, maybe, in the services that it is going to provide. Hardware requirements like network throughput are key for these applications. Which was another question I had; what applications will be provided by the NFV? These could be simple network services like DNS and DHCP, but also firewalls, loadbalancers and CDN like services to get OTT providers closer to the end-user. Other solutions that could be developed are architectures that centralize the intelligence and use cheaper devices for the end-customer that are easier and cheaper to manage. But it is clear that this is still in the conceptual status.
We also learned more about the possible business models for SDN and how they can close the business case. Price alone is not te differentiator. The possibility to create new services and slice the network are the real game changers. Running the network cheaper and more efficient is the second advantage. The possibility to decouple the hardware from the software creates the possibility to commoditize the hardware. This will change the supply chain. Vendors are being asked (forced) to get more innovative and work on integrated solutions with the providers.
Decoupling the hardware from the software makes it much easier to develop software and iterate the network. Openflow switches give the added possibility to test the software in an emulated network and create what-if scenarios.
If there is on big conclusion to take from this day it is that SDN and NFV is not about technology. Business drives these solutions and even than it is impossible to integrate these solutions without changing the people and the culture first.
The last thing that sounded very interesting to me was a proposal that Huawei is doing: Protocol Oblivious Forwarding makes it possible to program the network matching on everything in the packet or frame where Openflow can match on IP alone. I still have to investigate this and I also have to find the usecase for this, but it is certainly an interesting proposal.