Sometimes it seems like the world has gone crazy over software, no less in networking. Why? Putting aside all the VC driven reasons (low-cost of funding, higher valuations, difficulty in getting hardware shelf space in consumer markets,..), let’s focus on two: agility and spectrum of scale. Of the two, the latter is the most significant, and the area still left that will reshape networking as we know it.
Agility is a bit of a tough one to tease apart, because with micro-coded ASICs and/or ASICs with microprocessor cores, a level of programmability and agility already exists. This has served some networking vendors really well when it comes to investment protection / support of new protocols, as was the case with VXLAN. While there may be many opportunities to make hardware even more fungible, let’s put that aside for a moment, because that is not the big picture issue of focus in this specific blog.
The important thing going on is the experimentation with compute / storage cluster scales that are larger than a single CPU complex on a switch or router, whether the cluster be on-prem software, or importantly, for the future, cloud-based software. The latter is extremely important to the future of networking, and why the subscription business model will also be so important in the future – more on business models in another blog.
What is important in any network, is to respect the diversity and capabilities of different nodes on the network. Some are doing extremely important, low-latency, real-time critical, local operations. Others are “lazily” sifting through terabytes of data, potentially from multiple networks / businesses, to find answers / recommendations that benefit every member of a service. There are also scenarios in between those two extremes.
One example was highlighted by Juniper Networks CEO, Rami Rahim, at Juniper’s NXTWORK conference this week. In this video, Rahim talks about how software makes scaleout architectures reliable and cost effective. It is a good point, both as intended, and also when you consider the other side of the that equation, what was necessary to create the total disruption, high-performance fixed form-factor switches – the building blocks. Looking at the total picture of disruption is always fascinating. That particular software dynamic is worth a blog of its own, but I mention it here, as just one example, with there being so many more.
Software is not a craze because it is inherently more important than hardware, a concept that makes as much sense as saying the mind is more important than the brain**, software is a craze because it is adaptable to so many different configurations of compute / storage, and we are currently experimenting with which ones add value to networking. Couple that with other IT trends like scalable storage, microservices, usage-based billing, and other enablers of either lower-cost entry or game changing architectures, and you have an environment primed for innovation.
One of the determinants of future winners and losers in networking will be which companies can develop compelling value propositions with new combinations of software and compute / storage, to solve problems that are not already solved, or not solved well by a small number of ASICs / single router engine. We see some of this in today’s controllers, analytics, and automation engines. However, it is reasonable to assume that the industry will continue to innovate, and there will be more to come.
When it comes to innovation and business models, unlocking new sources of value, has always been a good path to follow, and the industry needs to stay focused on that.
** Understood that in theory the “mind” could be abstracted from the brain and executed in any number of future environments, but for now, I am in the camp that believes brain structure, and individual differences in brain structure, impact the nature of the mind, as we know it, and of course, there is much we do not know. The only real point is that software executes on something.