FOCUS: The Open Networking Foundation (ONF) promotes software-defined networking (SDN) through open standards. Where is the ONF’s stance on the current industry initiatives around standardizing  network protocols? 
The ONF does standards where it makes sense, such as where there is a protocol between two disparate pieces of equipment that requires a standard – especially when it is implemented in hardware. This is why we are standardizing the OpenFlow product – our standard communications interface managed by us. OpenFlow serves a basic purpose and we believe it is not where you need to innovate. You need to innovate below it on the transmission technology and connectivity, and above it in terms of control, and far above that in the applications.

The networking industry has been fixated on protocol standards for 30 years. But networking is becoming an element of compute, so standards mean something entirely different now. When it is a matter of software, if you don’t like it you can change a few lines of code to update it in real time. You can then have an open source community play with it. You don’t need a committee to decide in advance what it’s going to do, you should let the market figure that out for itself. The ONF is fostering this approach where it makes sense – further in the control and orchestration layers. We have been reluctant to standardize any interfaces and software.

People who want to write applications to a network infrastructure that’s abstracted, want to know what to write it to. This doesn’t mean there has to be just one interface but you don’t want 30. This is the approach we are taking to the Northbound interface. We resisted calls for us to standardize this but are exploring it and writing about it, and prototyping some examples of interfaces for the industry.

When it comes to efforts such as the Linux collaborative’s OpenDaylight, people say there is a standards war. But OpenDaylight does not do standards, it is doing substantiation of the OpenFlow protocol standard and all this other stuff above it. Hopefully, this substantiation will serve some parts of the market and serve it well, but it won’t serve all parts. If this substantiated standard becomes a popular piece of software, a lot of companies will build support around it and then it will become a convention.

We are interested if these interfaces – whether protocol or programming – are open and published, and not controlled by a single party, so people trust that if they build to them they are not going to have a company pull the rug from underneath. We want to see if the industry as a whole is contributing, and that changes made are merit based and controlled by a community.

So are you seeing more interest in OpenDaylight now?
OpenDaylight is gaining traction among developers and its co-contributors are also growing a lot, but the diversity of committers is not very high yet. I hope that will change. I haven’t seen any traction with network operators, but I am sure we will find some. Maybe it will be an enterprise. If enough people contribute to OpenDaylight and it is tested enough I am sure it will find a home. We are working closely with them. In March, we commissioned a development of open source that runs on OpenDaylight. 

We look at OpenDaylight as a useful industry effort to build tools that go on top of OpenFlow and build on SDN, and that’s a good thing.

It seems all major vendors have an SDN strategy. But I get the feeling there is some discrepancy about how concrete these strategies are and how they actually comply to the definition of SDN...
You do get companies that will call something SDN regardless of what it is. Virtualization and SDN are not the same thing but virtualization is a popular use of SDN. The latter doesn’t require network virtualization but it does require an abstraction of the network, and this is built into the OpenFlow model.

SDN abstracts the forwarding plane and the external control to it from a logically central place with a programming interface which allows the network operator to directly incorporate network policy.

This brings me on to Network Functions Virtualization (NFV), which the ONF has taken considerable interest in of late. How is this tying in with SDN?

NFV is a movement that the carriers have started. They love it because it gives them some medium- and short-term benefits. The problem they are trying to solve with NFV is they have single function appliances that are expensive and numerous. In some networks, the number of appliances is as great as the number of switches and routers. Switches and routers are expensive and each type of appliance has its own team of administrators.

The carriers want to convert these to software routines in commodity servers or maybe just single or multiple VMs. I think this will solve some of the appliance problems and give VMs a back door to using SDN and to have a single software approach for operating infrastructure.

Look at the most popular virtualized network functions such as load balancing and firewalls. It is possible to do some of those in a server. But there are not many you can actually do without getting the help of a forwarding plane – a connection that takes queries as they come in and sends them to a server for load balancing software routines. It says ‘here is the function you are after’, ‘what is the least congested server and least congested path to getting there using the OpenFlow switches?’. It is a rudimentary use of OpenFlow but without something like SDN and OpenFlow underneath you can’t really make this a virtualized function.

So our role in helping NFV succeed is to put in mechanisms to enact the least important virtual network functions. We want to see NFV succeed because it is going to be helpful to the operators, even if they just think about it as converting some of these things to computing tasks. But what comes with it, under the covers, is all the benefits of SDN.

The market for NFV seems to be moving quite fast for the vendors and the user community…

We are thrilled that vendors are putting stuff out. It means we are getting experience and feedback from the market, and that’s the most valuable thing there is. You can start now, do things gradually, migrate and experiment. There are some basic products and services, and they will get better over time. But you don’t have to wait for the ultimate solution.

I was on a panel in December in Korea with some CTOs. One was saying he couldn’t see the whole package completed yet. He asked me how I will compel him to adopt this, and I said ‘I’m not going to compel you at all but your competitors will because they’ll offer services and features at prices and rollout rates that you won’t be able to match’. This was a Tier 1 provider.

How is this move towards NFV going to affect the telco and service provider market then?
There is a lot of diversity among these telecom operators in their thinking and DNA. What are they really good at? How much of a software mind set do they have, and how many software skills? You are going to see a shakeout of networking operators over the next ten years. You will see a lot more virtual network operators of different types that specialize in their layer. They will book services from operators below them – real or virtual – and sell services to those above them. I think some operators that try to be everything won’t be able to in future. They will have to be more focused, specialized and smaller.

I think the network is going to be just a layer of accessing cloud data centers, and that’s where more enterprises are going to get their IT services. They are going to have less of their own inhouse networking staff because they don’t need to do big internal IT, and you will instead see more applications people conducting the business priorities to the infrastructure. This will take a while to happen but this is the direction it has to take.

Are we well on our way to this vision?
The vendors are getting vocal but operators and network users are not so much. A lot of them, however, are doing trials. Smaller enterprises are willing to talk about doing SDN and NFV. The hard part is getting customers to talk about what they are doing because a lot of them are doing small-scale trials and they are looking at these efforts to date as a competitive advantage. Some major telco operators are now talking about it, some financial services companies and some enterprises – but not that many.

I wouldn’t say there is any maturity of adoption, but I think that more IT managers understand that this is something they are going to have to deal with, and they are starting to ask vendors what they can do. The world is changing and software is going to be king of the realm. This will take some adjustment in business models and it will be a long process. I think the hardest thing to overcome will be organizational inertia regarding the way organizations are structured, how people are deployed and the skill sets that will be valued. These will all change a lot over the next ten years.

A year from now I think we will have some deployments of NFV based on SDN, and I think that we will get more enterprises and operators that can talk publicly about what they have done with SDN because they will have seen success.

This article first appeared in FOCUS issue 35. To read the full digital edition, click here. Or download a copy for the iPad from DCDFocus.