Too costly, too complex?

One of the key drivers of developments in networking hardware is software-defined networking (SDN), an emerging architecture that's being taken very seriously by some of the major network vendors, which are pitching it at their largest customers.

From a technical standpoint, SDN separates the control plane from the data plane in network switches and routers. As a result, the control plane is implemented in software, in servers separate from the network, and the data plane is implemented in 'commoditised' network equipment.

SDN will increasingly be seen as a solution for organisations with multiple data centres sporting the largest of cheque books.

Andy Robb is chief technology officer of Duxbury Networking.

In effect, SDN removes the 'intelligence' from the network switches and routers, and hands it to software programs, which then define the best paths for forwarded data packets.

This decoupling allows for the control plane to be implemented using a different distribution model than the data plane. And it allows the control plane development and runtime environment to be on a different platform than the management CPUs traditionally found on hardware switches and routers.

In order for the control plane to communicate with the data plane, the OpenFlow standard interface is adopted for configuring, managing and controlling switches and routers. Because the data plane - the hardware layer - is commoditised, it has opened the door for a range of third-party applications, from a number of industry players.

While many vendors have thrown their weight behind SDN, largely because it breaks new ground in terms of the way data transport networks function, other vendors are sceptical. They question how (and if) smaller organisations will accommodate SDN.

My belief is that SDN, requiring a range of open tools and software applications, does nothing to make networks simpler or commoditise the solution. Consequently, it will become too costly, too complex, and ultimately unsustainable in the networking space occupied by average-sized enterprises. In addition, mainstream enterprise adoption will require more enhancements and standardisation from SDN developers, which could take some time to achieve.

While the vendor agnosticism of the SDN model seems, on the face of it, to be positive, there is a lot to be said for leveraging existing topology protocols in many of the current platforms, which are just as capable of achieving SDN's goals - at a fraction of the cost.

Let's face it, current network technology has resolved most of the pressing problems associated with scalable LANs/WANs, routing and redundancy, while security, speed and bandwidth issues are in hand.

Many of the tried and tested platforms already provide centralised visibility and control over the entire network, at the same time improving application delivery for dynamic environments leveraging virtualisation, public and private clouds, server/storage consolidation and the commoditisation of IT.

What's more, current platforms are able to rapidly provision and de-provision servers and network infrastructures to ensure operational efficiency and the reliability of application delivery.

While SDN protagonists are doing their best to evangelise SDN as 'the' open source solution, many current unified management platforms provide support for the myriad third-party network devices already powering existing networks and integrating virtual data centres.

SDN supporters point to the problems associated with manual VLAN (virtual LAN) provisioning on conventional networks, claiming that SDN would cut time and reduce effort in this regard. In fact, the maximum number of VLANs available in most large-scale networks has already been allocated, and 'workarounds' often have to be initiated that SDN technology would struggle to address effectively.

So, is SDN technology the 'silver bullet' aiming to kill off the manufacturers of routers, switches and other conventional networking equipment? If it is, it will have to offer new levels of innovation, vision and a cost-effective solution to the managers of average-sized networks. The question is... will it ever?