Technical platform engineers at T-Mobile US are increasingly confident the operator can run all of its IT workloads on Kubernetes.

“We see Kubernetes as our platform forward for the future and we will continue to drive more and more of our application landscape to Kubernetes as time goes by,” Joe Searcy, a member of T-Mobile US’ technical staff on platform engineering, said during a webinar sponsored by Portworx. “We don't see a lot of reasons not to put everything on Kubernetes at this point.”

T-Mobile US began using Kubernetes about three years ago and today has a team of about 10 engineers, most of whom have traditional infrastructure backgrounds, working on the containerization platform. “Now, within the span of a few years we’re around 90, almost to 100 clusters, we have hundreds of namespaces, thousands of nodes, tens of thousands of pods across our infrastructure,” Searcy said. 

“We’ve been able to manage that with a really small team” that leans on the “declarative nature of Kubernetes itself to drive a lot of these pieces,” he said. Those pieces, including container image registry, ingress controllers, logging, monitoring, service mesh, persistent storage, dynamic tracing, and traffic steering all “comes out of the box when we deliver Kubernetes environment and we have a consistent experience between what we build on-premises and what we build on the public clouds,” Searcy explained.

T-Mobile US’ front-end APIs and web apps, pieces of the billing system, identity and fraud management APIs, systems that manage equipment agreements with devices, and point-of-sale processes across mobile, in-store retail, and online are all running on Kubernetes, he said. “Things related to how we plan our actual wireless network infrastructure, where we put towers and things of that nature, that all runs on Kubernetes” too, he added.

Because of the declarative nature of Kubernetes, T-Mobile US’ heightened degree of confidence that changes will be implemented correctly, and the technology allows engineers to break up code and deploy smaller chunks of new code and features more frequently, the carrier suffers fewer outages and when it does run into problems it’s been able to overcome those more quickly, according to Searcy.

The operator has, for almost two years now, been running more containers than virtual machines (VMs), and that container-based platform has allowed the company to lower its compute costs in kind, Thom McCann, a senior management at T-Mobile US who leads the carrier's cloud efforts, said during a presentation in October.

The operator is changing code dozens of times a day now, he said. “It’s way easier to deploy updates. We can now take an idea that our executive team or our marketing team has at the beginning of the day and have that ship out in production, across all of our data centers and cloud providers, by the end of the day, sometimes probably even within hours,” McCann explained.

The operator no longer has to wait for cabling of new physical systems in a data center or worry about access to adequate power. “We just kind of click some buttons, so to say, and it happens,” Searcy said. 

“My background is in that traditional infrastructure — Unix and Linux systems engineering. Kubernetes and these types of capabilities are the things I’ve been dreaming about for years. It takes a lot of that effort and just really, really simplifies it,” he said. 

Prior to T-Mobile US’ use of Kubernetes, the operator, like most businesses, faced major burdens and it took a lot of effort, coordination, and bumps in the road to get new code delivered. “When we started this journey, and this goes back almost five years or so at this point, it took up to seven months and 72 steps to get a new line of code delivered in production,” Searcy said. “Just saying that out loud makes me kind of pause and think wow, like how far we’ve come and what we’ve been able to do now.”