By Scott Carey, UK Group Editor, InfoWorld

Introduction

Handing your Kubernetes cluster over to a managed service provider is like sending your kid off to college — scary at first, but ultimately a lot less work around the house.

The managed Kubernetes, or Kubernetes-as-a-service (KaaS), offerings from the three major public cloud providers — Amazon, Google, and Microsoft — have all made huge strides in recent years, helping customers run and orchestrate their containerized workloads without needing to know the ins and outs of YAML configuration files or worry about autoscaling, updates, and cluster management.

"When enterprises think about strategic questions, the initial inclination is to run it themselves. Then over time they realize that running it themselves gives them no competitive advantage, and it's more likely that a vendor can run it better than they can," said Stephen O'Grady, co-founder of developer-focused analyst firm RedMonk. "Is every enterprise going that route? Not yet, but the trajectory and direction seem clear."

Reduced Management Overhead

Sylvain Roy, SVP of Technology Platforms and Engineering at travel tech company Amadeus, put it simply: "There's less work, clearly. It's operated for us, which is important because we face a challenge where we need all the people to manage Kubernetes."

Similarly, a small team of engineers at construction company Strabag has been running containers since 2006 and transitioned to self-managed Docker and Kubernetes over the past four years. The group is now looking to automate as much cluster management as possible by modernizing existing applications and handing off the underlying Kubernetes cluster management to the public cloud.

Mario Kleinasser, head of cloud services at Strabag, said: "The process is about giving away what can be given away."

Andrei Rybka, head of compute infrastructure at Bloomberg, said: "When you don't have an SRE team or a team managing the Kubernetes release cycle, and you're focused on running applications and don't want to manage Kubernetes, leveraging a vendor makes a lot of sense."

Fewer Specialists Required

Kubernetes management skills are hard to come by and expensive, especially when you're writing your own YAML configuration files. If you have people who can manually tune Kubernetes clusters, you may want to hand off cluster management for more generic workloads to free them up to manage your internal platform or any particularly important or tricky workloads. "Acquiring and retaining talent for these technologies is not easy, and that's clearly a challenge," said Amadeus's Roy.

Better Reliability

Simply put, large cloud vendors are generally better positioned to manage your Kubernetes clusters than you are yourself, due to the scale of their engineering teams, their broad visibility into customer deployments, and their access to underlying telemetry on those deployments.

"The vendor is likely to run it better," said RedMonk's O'Grady. "The vendor has telemetry and can see all customers running this system, which is different from a single enterprise that only has its own model to reference."

No More Upgrade and Patch Headaches

For those managing their own Kubernetes, upgrades and patches are among the least enviable tasks — which is why managed providers prioritize taking these off your plate.

Deepak Singh, VP of Compute Services at AWS, said: "Patching, updating, and managing Kubernetes yourself is a very complex thing to do, and it's completely undifferentiated heavy lifting."

It's Still Open Source and Portable

Managed providers must earn the trust of the open source community and customers who want to ensure the Kubernetes distribution they use stays as close as possible to the vanilla open source version, allowing for greater portability and avoiding lock-in.

AWS recently open-sourced its EKS distribution on GitHub to demonstrate this. Kubernetes co-founder and VMware Tanzu chief engineer Joe Beda acknowledged: "It's hard to have this conversation without talking about lock-in. How likely are you to move? If you do, what's the cost? How much code rewriting and retraining is required? Anyone making these investments needs to understand those requirements, risks, and the tradeoffs."

So Why Isn't Everyone Doing It?

For large, complex organizations, there will always be some workloads that feel uncomfortable to hand off — whether due to sensitive data security concerns, tricky internal dependencies, or overprotective platform teams who want to manually tune their own clusters.

However, the reality is that all the reasons for running Kubernetes yourself are becoming less and less compelling. "When you see your peers doing it, that concern goes away, and you see more people realizing the benefits," said RedMonk's O'Grady.

Source: InfoWorld

Back to Tech Blog