Scaling Kubernetes without disrupting manufacturing

Sworn to secrecy

This organization is a global pharmaceutical and life sciences enterprise, operating large-scale research, manufacturing, and enterprise IT platforms worldwide.

This story is shared anonymously at the customer’s request — but it reflects a real-world example of how a global pharmaceutical company brought structure and consistency to Kubernetes operations, without slowing down business-critical manufacturing systems.

Identifying details have been kept confidential, but the challenges, decisions, and outcomes are real.

At a glance

• Industry: Pharmaceutical
• Footprint: Global operations across research, manufacturing, and enterprise IT environments
• Objective: Standardize Kubernetes operations at scale without slowing manufacturing
• Key use cases: Kubernetes platform for factory Manufacturing Execution Systems (MES), Enterprise Kubernetes-as-a-Service for internal application teams
• Solution: Spectro Cloud Palette, Dedicated SaaS
• Outcomes: Cluster build from weeks to minutes, slashing a whole year off a major project’s timeline

Kubernetes wasn’t the problem – managing it was

This pharmaceutical manufacturer prided itself on empowering developers across its different business units to innovate: by design, each team owned its own infrastructure. 

Kubernetes had been part of this organization’s technology landscape for years, and as its popularity grew, development teams had created clusters wherever their applications needed them — including in VMware environments and data centers that support core manufacturing operations.

By 2024, the platform team had identified more than 300 distinct Kubernetes clusters across the organization. Different teams had built and operated them in a variety of ways, often relying on manual processes. There was no single inventory, no shared approach to provisioning or upgrades, and limited visibility into day-to-day operations. That inconsistency also created security gaps, with required tooling not always in place. The platform team spent more and more time tracking down clusters and understanding how they were set up, instead of improving how they ran.

A MES modernization deadline brought this issue to a head. 

The company’s Manufacturing Execution System (MES) is provided by a third-party software vendor, which had switched from providing it as a traditional client-server app to a containerized service, which assumed a Kubernetes cluster to run in.

Because manufacturing deployments now depended on Kubernetes being ready first, infrastructure availability became part of the critical path. Supporting the MES modernization across the organization exposed the limits of the existing operating model, and made it clear that the team needed a more consistent way to run Kubernetes.

The shift: treating Kubernetes like a platform

The pharma’s platform team knew that managing clusters one at a time wouldn’t hold up as adoption expanded. It was time for a different approach, one that could balance each dev team’s precious speed and autonomy, with company-wide consistency and efficiency.

With that in mind, the platform team evaluated several Kubernetes management platforms, including Rancher, VMware Tanzu, and Kubermatic. 

As you might have guessed, Spectro Cloud Palette was the only platform that met every requirement. 

It allowed the central platform team to define how Kubernetes clusters should be built and operated, while still giving development teams the freedom to provision and run clusters within their own environments. By building on open standards like Cluster API and leaning into automation, Palette fit naturally into existing infrastructure practices without changing ownership models or established workflows. 

The human factor also proved decisive. From the first meeting, the Spectro Cloud team leaned in and built a strong technical relationship with the platform team that only deepened as the project continued — the collaboration remains extremely close two years later.

A whole year saved

With Palette in place, the platform team could shift from ad hoc cluster management to a repeatable operating model unified on a single management plane. 

They can now define approved cluster configurations once and reuse them across environments, while application teams can self-service provision clusters through the internal platform as needed. 

Operators also have centralized visibility of which clusters are out of date, and can push updates directly in a controlled way, saving time previously spent on manual follow-up and reducing the risk of clusters falling behind or missing security fixes.

But the biggest impact was the compounded time saved in building and scaling clusters across the enterprise.

Building a cluster from scratch used to take an engineer two weeks of effort, and the company used to estimate six to eight weeks from a standing start to get an application installed on new infrastructure and ready to go at a site. 

With Palette, it takes just 20 minutes to build the cluster, and the application can be fully ready to go in hours. 

The domino effect is staggering. Before Palette came on the scene, the company’s leadership had planned out more than a whole year of work to build out, migrate and upgrade clusters to run the new MES applications enterprise-wide. Palette made it possible in weeks, literally knocking a year of work off the project. 

“We made it so freaking fast we ended up saving a full year of time, and now we can spend that time improving automation and building out more sites.

It’s the square wheel analogy — once you take the time to put the round wheel on, everything moves a lot faster and works a lot better.”

— Container Product Owner

While the platform team previously had to plan far ahead to keep up with the company’s growth expectations, Palette now lets them respond as opportunities arise. When development teams ask to add more clusters, even in cases that weren’t planned for, they can be provisioned quickly.

What they’re building toward next

Against conventional wisdom, this manufacturer didn’t start with a small project, ‘quick wins’ or low-hanging fruit. It started with big, critical systems that had a lot of attention across the organization. 

The success they had with Palette was so dramatic that naturally plans are underway to expand this Kubernetes management model more broadly across enterprise systems as well as research and laboratory projects. They’ve begun working with Amazon EKS, using Palette to manage clusters consistently regardless of where they run.

And, like many enterprises, this manufacturer has been reassessing its VMware footprint in response to Broadcom’s pricing changes. The team is currently evaluating how Palette’s Virtual Machine Orchestrator (VMO) could run VM workloads alongside K8s on bare metal infrastructure, ultimately offering a path to reduce VMware dependency over time.

By turning Kubernetes into an accelerator instead of a bottleneck, the platform team can confidently support today’s manufacturing needs and keep pace as the business evolves.