The network today is more critical to the success of IT than ever before. As such, any disruptive change in networking has to be one that is assimilated into the production environment using a measured and carefully phased approach.
SDN is clearly a disruptive trend in networking. Traditional network devices such as switches and routers have a co-resident control plane and data , and as such, each device is effectively independent. With SDN, the control plane is now being centralized, thereby making the SDN deployment a more concerted, orchestrated and centralized solution. This is a significant shift that holds significant benefits in terms of operational simplicity.
However, in the grand scheme, we are early in the SDN cycle and the deployment challenges associated with making SDN mainstream, including areas such as security, resiliency and scale, are still in the process of being ironed out.
One area that is still quite nascent when it comes to SDN is the area of monitoring, troubleshooting, and instrumentation. The ability of tools to monitor and manage SDN deployments is evolving, and with it, the ability to troubleshoot, manage, and respond to network issues in real time. All of this points to the fact that the success of SDN will largely depend on the quality of the implementations, the support model behind those implementations and the commitment of vendors to invest in quality, scalable and enterprise or carrier class SDN implementations.
In parallel, we are seeing a big push towards cheaper bare metal and white box types of solutions leveraging merchant silicon. This, too, is a disruptive trend and one that has substantial benefits. Driving down costs, providing choice, and opening up the ecosystem of vendors is a powerful trend. What enables the move to white-box networking is the maturity and broad availability of merchant chips. Merchant silicon technology has been tuned and matured over the last decade and is now sufficiently suited for optimizing and indeed commoditizing traditional Ethernet networking.
In isolation, these are both powerful and empowering trends; SDN for the operational simplicity it brings to the table, white box technology for driving down cost and opening up an eco-system of vendors. However, what is interesting is that in many cases the industry is binding these two trends together, making use of white box solutions for SDN deployments.
This is a little worrisome. If history is any indicator, the adoption and maturing of a new disruptive technology or set of technologies such as SDN has typically preceded the commoditization of that technology. In other words, gaining a good understanding of a new technology, securing it, scaling it, and having the ability to manage and troubleshoot it, need to be resolved before the technology can be successfully commoditized. And while we are still in the early stages of SDN, the trend towards using bare-metal, white-box solutions for SDN seems to be abuzz.
That seems a little like putting the cart before the horse. Rather than allowing SDN technologies to mature and then cost-reduce and commoditize the solutions, there seems to be a push to commoditize the solutions in advance of the technology maturing. That could be risky. A poor implementation, a network gone wrong, and suddenly momentum could stall. I think de-risking SDN may imply taking a step back, perhaps, getting a firm grasp on what aspects of SDN are working, ensuring that the ability to secure, troubleshoot, monitor, and manage SDN deployments at scale are well understood, and then thinking of commoditizing the technology.
By the same token, leveraging merchant-silicon-based white-box solutions for traditional Ethernet switching makes sense. After all, that is a mature space and is well understood and ripe for cost optimization.
But white-box networking combined with SDN concurrently simply seems like taking on too much risk.