Best practices for Kubernetes use in the telco cloud from a telco cloud architect
Scalability is the essential characteristic of any cloud native-service: compute, storage, memory and other factors are deployed at scale, on demand. Telco cloud is no different, said Neil McRae. Except the scalable unit is the network itself. Neil is BT’s Group Chief Architect Managing Director, Architecture and Technology Strategy. Being cloud-native means more than just containerizing an app and orchestrating it via Kubernetes, he said.
“My team might disagree with me, but I kind of know what I’m talking about when it comes to building infrastructure,” he joked.
McRae sketched out some fundamentals for Kubernetes-based cloud-native telco infrastructure with Viavi CTO Sameh Yamany during a fireside chat at the RCR Wireless News Telco Cloud Forum 2022.
BT’s network is engineered to provide services at scale, said McRae. “Voice, TV, broadband, mobile,” McRae said. He described them as “the heart of everything that we do.”
“In the new cloud-native world we see many kinds of solutions for those four key…core capabilities, and we see them being treated as pets,” he said.
Network functions are cattle, not pets
What does animal husbandry have to do with the cloud? Pets get special treatment and a soft bed in the house to sleep on. Cattle and other livestock are moved en masse. Network functions need to be thought more like cattle, less like pets.
Applications ported to the cloud may only be partially native, McRae explained, which sharply limits their scale. App bottlenecks include vendor-locked deployment scripts, incomplete abstraction from infrastructure, and the decomposition of applications into functions rather than true cloud-native infrastructure. Partial automation is another roadblock.
“If we really want to think about cloud-native, the big driver…is around automation,” McRae said. The only way of building network functions simply won’t scale to the cloud-native world, he said.
The four pillars of cloud-native infrastructure
Exposure, security, lifecycle at scale, and interface and service mesh comprise the “four pillars” of BT’s Kubernetes app strategy within the telco cloud. Exposure starts with Application Programming Interface (API) gateways, access controls and fleet management.
“The world is moving to complete application abstraction,” said McRae.
Security is are crucial, especially given the scale of services, as are issues like hardening and regulatory compliance.
“You’ve got potentially billions of individual components on a cloud-native infrastructure,” he said.
The third pillar, lifecycle management, is where the build engine is managed.
“Observability is really important,” said McRae. “Is everything working the way it’s supposed to? And can you see it? If you can’t, how do you know it’s working?”
The fourth pillar, interface and service mesh, is also linked to observability, said McRae. “Different infrastructures have different interfaces, plug-ins to each other,” he said.
Containerization at scale demands consistency
Vendor-specific containerized apps can create multiple challenges for telcos starting with their security posture, restricted operational visibility, possible dependencies on other libraries and apps, and patch status.
“My challenge to vendors is around security and software version and making sure that every piece of software is running the latest patch level,” he said.
BT has its own process to make sure every app that touches the telco cloud passes muster.
“We put them through our own filter, if you like,” he said. BT signs all software images, and integrates with their own logging and monitoring tools.
“So we go from a land of pets to a world of cattle,” said McRae. “And we can manage those cattle really well.”
Ultimately all of this is a work in progress for BT and for everyone in the business. McRae acknowledges that BT’s journey to telco cloud is early on, and that some of foundational technology is still quite novel for its use.
“It’s early days. Kubernetes isn’t ten years old yet, and it’s already running the world.”
The post Three fundamentals of a cloud-native infrastructure strategy appeared first on RCR Wireless News.