SDN – the quiet revolution

Software defined networks

I first became seriously involved in software defined networks (SDNs) more than 10 years ago. At the time I was working with product and standards development teams influencing an ETSI policy document looking at SDN and network function virtualisation (NFV) as separate, but highly interdependent trends. The plan was to turn what was then a trial concept into a commercial reality.

While often mentioned in the same breath, SDN and NFV play different roles in redefining network architectures.

On the one hand, the main driver behind SDN is to separate the network control plane (essentially the routing that determines how data is sent from one place to another) from the transport layer. NFV, on the other hand, is about abstracting network functions from the hardware on which it runs. Theoretically, you could have SDN without NFV – a software defined network with static hardware.

The reason this work was – and remains – important is that service providers always want to get away from operational and architectural inflexibility and vendor lock in.

Before the concept of SDN, networks were hardware defined and the hardware market was dominated by two or three very big players that sold vertically integrated appliances that, while operationally robust, provided limited choice and offered very little operational flexibility. To address this, the mobile industry borrowed standards and interfaces like OpenFlow from the IT world to enable secure and efficient access to and manipulation of network routers and switches.

Today SDNs are becoming mainstream. It’s true that there have been some bumps along the way, not least because switch vendors and some of the big equipment suppliers felt they could make more money short term by maintaining the status quo. There is also the issue of squeezing the most out of (costly) legacy hardware. But the programmability of networks has become increasingly important to service providers. As data usage over mobile networks continues to surge, more granular resource management becomes a priority. The flexible, efficient end-to-end management of dataflows has become business critical to operators and SDNs hold one of the keys to making it happen.

SDNs mean that the management plane can be separated from the dataflow and that these aspects of network functionality can be sourced from different suppliers. This is partly about reducing capex by making it possible, for example, for operators to source hitherto expensive switches and routing gear from white box suppliers without comprising the integrity of network operations. But it is also about improving network efficiency by having more control through programable orchestration software.

It’s probably true to say that, as tech revolutions go, SDN has been reasonably successful. At some level, kit has become more generic and commoditised, but in practice significant levels of vendor lock-in remain. While the equipment will be standards compliant, there is usually enough complexity and sufficient proprietary features to prevent high levels of mix-and-matching delivering optimal outcomes. And the plug and play dream promised by the early SDN vision remains more aspiration than reality.

This said, SDN has enabled network operators to accomplish the vertical separation of control, user and management planes which, from an architectural perspective, is becoming increasingly important for the design and operation of 5G networks.

It’s tempting to equate the SDN trajectory with the ongoing architectural revolution implied by the current Open RAN initiatives. There are certainly legitimate parallels. Both have taken a similar amount of time to make the journey from drawing board to commercial reality. And both are likely to prove very much more difficult to implement in practice than seemed apparent at the beginning of the journey. But there are also some important differences. From the start, SDN benefitted from concepts and standard approaches in the IT networking sector and was shaped by objectives that were, in general, universally agreed. Open RAN is much more fragmented – not just from the point of view of objectives and standards, but also in terms of the political and regional agendas informing and, in some cases, inhibiting progress and consensus.

From the Real Wireless perspective, we believe that the SDN revolution has been a progressive evolution of network architecture because it provides more resource control to network operators and more supply chain diversity, which is always something we champion. The other side of the diversity coin is complexity and this is the area in which we support clients – helping to guide technology choices not just to address current needs, but also to protect investments on the path to 5G and beyond.