If I asked you to close your eyes and name all the little icons that show up in the upper right corner of your phone screen at a given time, would you be able to? I’m talking about the 5G, LTE, Wi-Fi, Bluetooth, GPS icons—all those little symbols that tell you which network(s) you’re currently connected to. I’d wager most people couldn’t do it. And that should say something to communication service providers (CSPs), vendors, and all of us who spend so much time thinking about network technologies: customers don’t care.
It’s not that those different access types don’t matter, or provide important capabilities to connect our digital world—they absolutely do. It’s just that we shouldn’t expect customers to think about them. To both consumer and enterprise users, networks just get them to the content and applications they care about, and the less they have to think about them, the better. In a perfect world, users would just get the best connection for whatever they happen to be doing automatically, without having to think about it at all. Figuring out how that happens—reliably, securely, with the best possible performance—is our job.
In the short term, enterprises will still need to navigate different access networks, each with its own specialized technologies, standards, and tools. But the future promises something different. We’ll still use many networks, but all the complexity in deciding which one best suits which application, under which circumstances, will get abstracted away. An AI-driven overlay will choose from among all those networks, dynamically optimizing connectivity without anyone having to tell it how. The same overlay intelligence will flow the other way too, exposing all a CSP’s diverse multi-access and edge capabilities through an open API, so that applications can program exactly the connectivity they need.
That’s the kind of dynamic, programmable connectivity that’s possible with convergence. And it’s going to transform digital experiences for everyone.
Reimagining Convergence
There are lots of ways to converge underlying technologies to create a better customer experience, and over the years, CSPs have implemented many of them. Triple- and quad-play offers, for example, are basically just billing convergence. By combining disparate services—TV, Internet, cellular—into a single purchasing experience, sometimes via a mobile virtual network operator (MVNO), they become much easier for customers to consume.
But it’s not just about billing. As the industry advances, we see convergence also making strides in the technical landscape. The underlying infrastructure is now aimed at making connections seamless, regardless of the medium. For customers, the end result is a hassle-free experience; for enterprises and IT teams, it’s about efficient and agile network management.
Enter the world of access networks. While they may seem like a technical realm far removed from the average customer’s daily experience, their convergence is essential for ensuring that said experience remains uninterrupted and high-quality. Modern software-defined wide-area networks (SD-WANs), for instance, can switch to cellular backup if primary connectivity fails, or steer non-critical traffic to less expensive broadband links. Recent 3GPP releases add new convergence capabilities to 5G, such as switching between private cellular networks and Wi-Fi to offload data and enable seamless user experiences. Many operators also now work with low-Earth-orbit (LEO) satellite partners as a cost-effective alternative to building new cell towers, while maintaining seamless connectivity for users.
Now though, it’s time to take the next step, where convergence doesn’t just create a better customer experience but unlocks transformative new use cases. Those include specialized network slices for self-driving vehicles, automated factories, augmented reality and virtual reality (AR/VR) gaming, and others that apply the right connectivity to meet the requirements of each application. But that’s just the beginning. Imagine networks that calculate the best resource for each request, from the best edge or cloud location, to generate the least carbon emissions, or achieve the lowest cost-per-bit, on a moment-by-moment basis. The kind of convergence I’m describing will empower customers to program networks to optimize for practically anything, or many things at once.
Building Overlay Intelligence
Most networks—cable or fiber wireline, 5G cellular, Wi-Fi, what have you—already employ some degree of intelligence. Modern telco networks, for example, support traffic engineering and quality-of-service (QoS) prioritization. SDN edge intelligent controller and radio access network (RAN) intelligent controller (RIC) platforms add still more programmability to business networks and the telco RAN. These are important advances, but they still largely work from the network in, programming specific behaviors for each underlay. What’s missing for more transformative use cases is an intelligent overlay.
CSPs and their customers need an intelligence that lives on top of the disparate underlay options, that can program all those domain-specific controllers and optimization mechanisms, but has access to the application layer as well. Instead of just optimizing what the network provides to the service—analyzing things like signal strengths, network KPIs, traffic engineering policy—this intelligence can make decisions based on what applications need from the network, and execute from the top down.
Such an overlay intelligence would continually analyze all underlying networks to provide applications with the best possible connectivity at any given time. But it would also abstract away the technical complexities of those underlays, giving customers the ability to program how the converged network behaves using business-level logic, via an open API. Customers could specify things like, “Use the lowest-cost connection for my office applications, but make sure latency for factory equipment never exceeds 40 milliseconds.” No one would need to code out instructions to use this connectivity, with these configurations, in these circumstances. The intelligent overlay would just do it.
This open and converged, programmable network will be:
- Dynamic: Today, it’s up to you to figure out which resources to deploy at which network and edge locations to support upcoming customer applications, and hope your capacity planning works out. Tomorrow’s network will function more like an all-encompassing collection of flexible resources—some in central data centers, some in edge clouds, some at customer premises. It will automatically determine the best way to get each application what it needs, and continually optimize those decisions as network conditions change.
- AI-driven: Current network controllers can perform a rudimentary version of programmable connectivity, such as switching networks when throughput dips below a certain threshold. But that’s one input to consider when deciding among two or three network options. What happens when the number of inputs and decision points increases a hundredfold? What if enterprises want to account for throughput, latency, resiliency, security, energy consumption, data sovereignty, and more for every network request—across dozens of networks and thousands of edge clouds? No human can configure policy to cover every possible contingency, moment by moment as the network changes. But the task is tailor-made for AI.
- Edge-optimized: CSP edge clouds are among the most exciting new opportunities, giving enterprises powerful new tools to automate equipment, apply AI to real-time decisions, and more. But edge locations are inherently constrained; you can only squeeze in so much hardware. As more applications depend on those resources, you’re looking at expensive ongoing truck rolls and, if you’re not careful, new hardware silos that make the network less efficient and harder to manage. Tomorrow’s converged, intelligent networks will understand what each application requires and automatically deliver it in the most efficient way—across all available resources in edge clouds, data centers, the 5G RAN, and beyond. So your edge investments will last longer, even as you scale up support for new applications.
I’ve been focusing on access networks, but CSPs will be able to extend this intelligent programmability to any resource in their environment—RAN, core, edge, satellite, cloud—reassembling them on the fly to meet different application needs. Effectively, you’ll give customers the ability to program their own custom network slices. And you’ll transform your network into a multipurpose platform for third-party integration and innovation, with connectivity that’s optimized, scalable, and on-demand.
Looking Ahead
At VMware, we’re taking steps to make this vision of converged, programmable connectivity a reality. We’re working with CSPs and the broader ecosystem to open up network interfaces with secure, open APIs, and harmonize disparate underlay resources. These efforts build on steps that CSPs are already taking to make their networks more agile, automated, and programmable.
Now, by adding an intelligent overlay that can dynamically control those underlying resources based on simple, business-level directives, we’re building something truly transformational. Suddenly, CSPs will be able to create new customized services in minutes. They’ll empower enterprises and third-party developers to continually roll out innovative new use cases that take advantage of everything the converged network has to offer. And they’ll do all of it with a network that continually optimizes itself to be more secure, reliable, and sustainable.
