Automation and Big Networks: The More, The Better
Recently I had the good fortune to be invited to address a Light Reading BCE audience in Austin, where the theme was the future of big networks. There was plenty of discussion, as you might imagine, about software defined networks/SDNs, network function virtualization (NFV) and artificial intelligence (AI) — all of which matter greatly to the future of big networks.
But for several reasons, if I had to pick one category of work that best represents the future of the network, it’s automation.
Why, because it’s the only way to keep up with the ceaseless changes that come with marketplace disruptions. Such disruptions are happening in every market segment, really, not just big networks — think about how our children are learning, compared to how we learned. Or how we bank: 95 percent of Sweden is cashless, now, and Denmark wants to go entirely cashless. The ATM machines that were such a game changer in the 1970s, are now being removed by the thousands.
And here’s a particularly sobering prediction, from IMD: In three years, 40 percent of the incumbent players in our market aren’t going to exist anymore.
These disruptions are happening because of digitization, which is at the root of everything — including us. The good news is, broadband consumption continues to grow 50+ percent every year, with no real hint of stopping – (to be exact, Cisco VNI data says that globally, peak Internet traffic grew 51% in 2016). What a great business to be in, where every year grows!
The bad news is, revenues aren’t growing at nearly the same rate — which means we have to be creative. We need to get exponentially more done, in exponentially (or, at least, historically) less time.
As a mathematician, I appreciate the challenge that is a double exponential: More cycles, in less time. The answer can only be in automation, to be faster and more efficient, with ruthless simplification of our networks, and relentless collaboration.
The tools are there, in SDN, and our collective use of the clouds. And in programmability, certification, virtualization, open source — it’s not one or the other of such tools, it’s all of them together. Tools matter to how we simplify our networks.
Here’s what I mean by that, tactically: The days of using a command line interface to manage a box are going away. Why, because it’s not about managing boxes anymore. It’s about managing networks, with model-driven networking. Taking that piece of software, that used to be in a box, and moving it to a higher-level control plane — to be able to look at the network, and manipulate it, more holistically.
Which brings us to standards. We’re involved in so many standards organizations, and open source associations — the whole “so many to pick from!” quip isn’t so much a joke anymore. You’ve probably asked yourself this, too: Are they really open? Any time any of us has to do the same work four times, it means that we could’ve used that energy elsewhere, to move something forward in a more efficient way.
The more we can simplify and unify, the more we can innovate. By that I mean getting all the way into deep-level data to use in analytical ways. Leveraging artificial intelligence/AI, to cross-correlate what’s happening on the optical link, or the firewall, or the router, to make intelligent decisions.
For web-scale companies that grew up on broadband, automation is already a big thing. They’re 100 percent automated; “the box”is abstracted . By contrast, for traditional service providers, the box means everything! Each component requires massive fault tolerance, with every possible backup failure plan — which makes for a very box-heavy network. For web-scale players, none of that matters, because it’s all one layer up. Something fails? No worries. Something else can automatically take over, because it’s all in that external level of control plane.
There’s another very real reason to automate, and that’s the amount of operational costs that can be removed. We’re already seeing this in data centers — yes, someone who runs a data center is far different from someone who manages an entire outdoor infrastructure, but there’s something to be learned there about what automation can do.
The first step in the journey to automated, holistic networks is orchestration. Orchestration is the foundational step to launching new services, running them, and monitoring them. If you haven’t done it yet, you really should!
And then there’s the very hot topic that is compute power: Figuring out where we put tens, or hundreds, or thousands of boxes; finding the perfect allocation for mobile edge resources; establishing where deep-edge compute goes. The whole mobile edge discussion, by the way, reminds me a lot of CDNs, except that was primarily for video caching. We didn’t really lead that charge, as an industry — the CDN providers did. With mobile edge compute, we have a real opportunity to take the lead, and establish its location, use, and business rules.
As in the keynote that triggered this blog, I’ll end with a call to action: Let’s figure out what we have to do, as an industry, together, to get automated. What are the right open standards and open source components that’ll get us there — at scale? Our collective network scale is phenomenal.
It’s my belief, and Cisco’s belief, that there’s never been a better time to look at how to automate so much of what we’ve done manually, in order to increase our speed of innovation.