The arrival of 5G brings with it new levels of integrated infrastructure and innovative cloud services. But, to make the most of its possibilities, the industry must deliver many more applications at the edge of the network. In this fast, low latency new world, network operators need to standardize application deployment across clouds, data centers and at countless edge locations, all at unprecedented scale.
Enter Deutsche Telekom. The first global Telco running Kubernetes across both its core and edge sites to deliver multiple use cases through a single, telco-grade platform, the company’s collaboration with Weaveworks marks a milestone in the mobile industry. Dubbed Liquid Metal, it is a fully open, GitOps-enabled, distributed Kubernetes Cluster-as-a-Service (CaaS) platform, operating at massive scale. It is already empowering thousands of application owners and vendors to securely deploy workloads and other services, on-premise and on either VMs or bare-metal servers.
Why Kubernetes and GitOps for 5G at the edge?
Reducing latency is just the beginning
The promise of 5G has already sent the expectations of mobile users skyward. But to deliver on the protocol’s promise of transformative speed, radical change is needed behind the scenes. Levels of latency that were once acceptable will render many 5G applications
impossible to deliver. The solution is to move compute resources nearer to the edge of the network – and therefore, closer to the end user. But there’s a problem. When compute moves out of the core data centers, complexity increases rapidly, because suddenly, every single cell tower is a location that needs to be managed.
A consistent, declarative platform across thousands of locations
Kubernetes is a cloud-native platform for managing containerized services. But what makes it special is its declarative nature – and its associated ability to automate much of the hard work of keeping things running. At extreme scale,this ability to automate away much of the complexity is critical. However, Kubernetes was designed to run in the virtualized world of the cloud. When it comes to delivering 5G, that’s a limitation.
5G depends on bare metal
Specialist hardware is required for a radio access network, such as antennae. And to access this specialist hardware, it is often necessary to avoid the abstraction of VMs and instead run on bare metal hosts. For Deutsche Telekom, this presented a problem. On the one hand, they needed to shift compute resources out of the datacentre and into thousands of edge locations. But on the other hand, they needed to run on bare metal – which appeared to rule out Kubernetes, the best platform for operating at scale. The solution was Liquid Metal – a bespoke infrastructure platform built using Weave GitOps, to dynamically create and manage Kubernetes clusters on both VMs and bare metal hosts.
The edge demands extreme efficiency
Moving compute to the edge brings all kinds of benefits - reducing backhaul and making the network more efficient overall. At the edge of the network, however, compute is a scarce resource. Unlike huge datacentres, edge locations have limited space, power supply and cooling facilities. The resources available must therefore be used as efficiently as possible. Liquid Metal is such a lightweight solution that it enables Deutsche Telekom to run more clusters than would otherwise be possible on the limited pool of physical machines. It conserves resources further by running the Kubernetes control plane on lightweight microVM technology (such as AWS Firecracker) which enables Kubernetes hosts to be completely booted in as little as 10 seconds. Worker nodes that require hardware access, meanwhile, run on the bare metal.
One model to manage it all
So far, so good. The workloads had been moved to the edge, where Kubernetes was now able to manage them on a combination of bare metal servers and microVMs. But the essential problem remained: having thousands more locations to monitor and manage. Deutsche Telekom needed a way to stay on top of what was now a wildly distributed infrastructure. Without GitOps, it would have been impossible. (Learn more in the press release.)
In 2019 we saw that Kubernetes and containers would transform how we deliver software inside telcos. By using best of breed open source and selected leading contributors to cloud native technology as partners, we could enable a rapid 5G rollout,” said Abdu Mudesir, SVP Technology at Deutsche Telekom, “As soon as we empowered our internal Kubernetes platform team to move forward with Weaveworks’ support, they shipped a working system into production in a matter of months, and onto 5G in 2021. All this has been possible through the GitOps model
How GitOps can scale platforms and applications without scaling teams
Weave GitOps gives platform operators the ability to go beyond Kubernetes-as-a-Service and create a tailored Platform-as-a-Service, using profiles (profiles are defined in catalogues, providing a set of approved applications that can be installed on clusters at creation time or subsequently). Profiles enable operations teams to define cluster components such as Observability (as an example you could have Prometheus and a Grafana stack) as part of your ‘Dev’ or ‘Production’ cluster definition.
Cluster lifecycle management
GitOps makes it easy to manage the full cluster lifecycle, at massive scale. A small team can create, update and delete clusters across whole fleets, with every action recorded in version control. This means practically instantaneous rollbacks and complete observability at all times. In short, it makes things much, much harder to break.
Built for a multi-cloud world
On-premise and vSphere, Amazon EKS or Azure AKS – the reality today is that all large enterprises are multi-cloud, running their platforms across a range of different infrastructure providers. Weave GitOps was designed to support platform and fleet management across multiple infrastructure providers.
Liquid metal for hardware access
Liquid Metal enables Kubernetes to run bare metal, so it can access hardware features directly, such as Radio Access Networks (RAN) for 5G. Kubernetes clusters can be provisioned dynamically on both bare metal and lightweight virtualized nodes (microVMs). It was developed by Weaveworks and Deutsche Telekom, specifically for high performance cases where hardware acceleration is needed, such as 5G.
Continuous application delivery
It's not just about the platform. It's about the applications that run on it – and in the case of Deutsche Telekom, deliver the benefits of 5G to millions of devices and customers. Continuous application delivery involves rolling out updates across thousands of locations and fleets of clusters. GitOps makes that scale possible – but crucially, without scaling your team in parallel. What’s more, the confidence it inspires in delivery pipelines means teams can deploy more frequently, providing better services and quality to end users.
Developer focused interfaces
Different stages in the workflow require different interfaces. The graphical user interface provides a single pane of glass for the fleet of cluster and applications, providing observability through dashboards and visual cues. The CLI provides developer focused
tooling when precision and repeatability is required. Automation can also be achieved with our CLI or APIs.
“GitOps has helped to make Deutsche Telekom’s containerized clusters easier to manage and more responsive. They require less human management intervention. The team’s ROI can be judged by the scale of its achievements—managing thousands of clusters with a tiny team, and managing cloud development and cloud operations continuously.” Jon Collins, Analyst, GigaOm
Go further (and faster) with GitOps
GitOps is the path to automate Kubernetes and build robust Kubernetes platforms across any infrastructure and at any scale, even telco scale. Continuous delivery, deployment and control are the first step in a successful cloud native journey, followed by progressive delivery and self service capabilities. Weave GitOps is the modern operations product which enables this transformation – so contact us for a demo today.