Convergence has been buzzword in the telco world for decades now, ever since phone companies transitioned from switched networks to IP-based ones. Today, Alexander Graham Bell would hardly recognize the massive telco/media conglomerates that descend from his venerable Ma Bell.
And yet, while it’s true that voice/Internet convergence has enabled us to watch videos on our smartphones, there remains one holy grail of convergence that has yet been out of reach of the telcos.
The Trends that are Disrupting the Status Quo
The pinnacle of modern telephony today is 4G LTE – a set of protocols that empowers the mobile telephony revolution, penetrating the daily lives of people around the globe. But there’s still something missing.
4G LTE connects our phones, as well as a range of other devices, from tablets to home alarm systems. But from the enterprise perspective, the telcos are simply providing communications services – services that stop at the smartphone interface.
5G promises to change this equation, bringing the long-desired convergence between telco and enterprise services.
While it’s true that 5G is a next-generation replacement for 4G LTE, bringing astounding performance to our mobile devices, it’s much more than that. 5G is actually a family of protocols that operate at different distances.
An oversimplified, but useful way of thinking about 5G is that it combines replacements of 4G LTE, Wi-Fi, and Bluetooth. (This representation is an oversimplification largely because both Wi-Fi and Bluetooth are undergoing their own next-generation transformations, promising to give 5G a run for its money – in particular, next-gen Wi-Fi with ranges in the tens of miles.)
Nevertheless, once 5G fully rolls out, the diversity of 5G-enabled endpoints will put 4G LTE to shame. Anything might be a 5G endpoint, from your laptop to the alarm sensors on your windows to the price tag on your clothing.
The Rise of the Edge Data Center
From the enterprise perspective, however, the most transformative aspect of 5G, even more than its impact on the IoT, is the role of the ‘mini’ data center.
As part of the 5G rollout currently in progress, the telcos are rapidly deploying small data centers in widespread locations around the globe, from cell tower base stations to community points of presence to office building server rooms. I consider such locations the ‘near edge’ to differentiate them from the endpoints at the ‘far edge.’
The ostensible reason that the telcos are investing so much money into these mini data centers is because the midrange 5G protocols (think next-gen Wi-Fi competitors) require geographically distributed base stations. For the telcos, however, the big win isn’t simply competing with next-gen Wi-Fi. The big money is in building edge computing capability for enterprise and web scale companies.
In fact, this 5G-driven buildout is making edge computing a reality where before it was largely smoke and mirrors.
Edge computing brings compute and storage resources closer to the consumers of those resources, lowering latency and improving the cost-effectiveness of compute and data-intensive tasks that require data from the far edge, namely AI training and inferencing.
After all, if you’re going to leverage AI to, say, uncover suspicious behavior in hundreds or thousands of video feeds from a large facility like a factory, power plant, or airport, you don’t want to send all those data all the way to the cloud. You want to process them locally – and the near edge mini-data centers are just the ticket.
The Missing Piece of the Puzzle
There remains a sizable gap in this story: the software infrastructure that must run in the edge data centers in order to support workloads like data-intensive AI applications, or any other business application that might run at least in part on the edge.
Today, that software infrastructure depends upon Kubernetes.
Kubernetes brings the massive scale and dynamic capacity that edge computing requires. In addition, many of these edge data centers run Kubernetes on OpenStack, essentially making them full-fledged ‘bare metal’ clouds in their own right, running predominantly on open source software.
Much of this software infrastructure isn’t important simply for its economics – it’s also important because open source allows organizations to extend projects to meet their diverse and dynamic business needs. True to form, lightweight versions of Kubernetes are coming to market, well-suited for running in edge data centers.
The word ‘lightweight,’ however, may be a misnomer. True, the open source community has crafted OpenStack/Kubernetes flavors that are well-suited for the tiny clouds that edge data centers are becoming – but there is nothing lightweight about the power of such software.
On the contrary: the community is also working to expand the scalability of Kubernetes to unheard of levels – Kubernetes deployments on the order of millions of clusters.
Why Do We Need Millions of Kubernetes Clusters Running Across the Edge?
Will millions of Kubernetes clusters running a commensurate number of workloads actually fit into an edge data center?
Perhaps, but that’s not the right question. The right question is whether we can abstract multiple edge data centers, as well as (traditional) clouds and on-premises environments in order to support millions of Kubernetes clusters across a complex hybrid IT landscape – where ‘hybrid’ is a mix of cloud, on-premises, and edge.
Today, running Kubernetes on top of such hybrid IT infrastructure is still quite difficult, but the writing is on the wall. Cloud-native computing will support effectively unlimited Kubernetes deployments running across clouds of all sizes and locations.
Unlimited scale is all fine and good, but given that today’s deployed Kubernetes environments extend to perhaps dozens of clusters at most makes one wonder what we’d ever do with millions of clusters.
The market as a whole and the ingenuity of its participants will be the final arbiter of such a question. But we can posit some likely scenarios.
Given a 5G-enabled IoT will generate vastly more raw data than we have today, it’s clear the AI processing we’ll want to do should be distributed across the edge.
Regardless of how inexpensive bandwidth becomes in a full-fledged 5G world, we’ll always have to deal with latency (thanks to that pesky speed of light). A widely distributed edge cloud environment will enable ongoing optimization of performance in the face of such physical limitations as well as exploding data sizes.
Web-scale applications will also continue to push the limits of available infrastructure.
We rely upon 100 Mbps+ home bandwidth, Wi-Fi, and public cloud infrastructure to deliver the over-the-top functionality of Netflix, Amazon Prime, and the others. 5G and a Kubernetes-empowered edge cloud will only redouble the efforts of such companies to deliver next-generation services to homes and businesses.
The real win, however, will be services that we have yet to conceive of – or at least, beyond the minds of a handful of visionaries.
We have hints about how such services will behave. Today we can run multiple versions of a new service to gauge the reaction of our customers. What if we could run millions of versions at a time, with a sophisticated AI algorithm making the decisions about which ones to scale up and which to scale down?
The Intellyx Take
If we’re running full-fledged clouds at every cell tower, then what does the word ‘cloud’ really mean? Should we continue to use the term, or will it lose significance?
When you peel away the marketing hype, a cloud is basically an abstraction of compute, storage, and networking resources. When we’re doing cloud computing, we no longer care about the specifics of how these resources work. We simply access them as services.
Given the convergence that 5G brings to the table, we can add communications to the mix. Now clouds abstract all four resources instead of the original three.
With the addition of edge computing, we might say we’re running clouds on the edge. Or perhaps well say that the edge is part of the cloud. But I think the best approach is to think more broadly in terms of cloud-native computing.
The essence of cloud-native computing is a comprehensive abstraction across all of IT that enables us to apply the characteristics of the cloud (like horizontal scalability, elasticity, and resilience) to the entire hybrid IT landscape in a consistent manner.
We can thus say that this entire story – traditional clouds plus edge clouds plus 5G plus all the endpoints – are all part of the cloud-native computing story. It doesn’t really matter if we still call anything a cloud.
© Intellyx LLC. Intellyx publishes the Intellyx Cloud-Native Computing Poster and advises business leaders and technology vendors on their digital transformation strategies. Intellyx retains editorial control over the content of this document. Image credit: Judy Schmidt.