Kubernetes Cluster Intelligence

When I joined CA Technologies, the FreshTracks engineering team had just created a prototype of a hexmap for visually navigating a Kubernetes cluster. It displayed tesselating hexagons for each level of the cluster hierarchy, and you could double click to dive one level further down, from namespace to workload to pod to container.

It was a really interactive and fun interface, but it wasn't quite ready for release. The cluster entities weren't labeled, and our user research showed that the available actions were hard for early adopters to discover. For the cluster size we were running internally, the map could handle the number of workloads, but it couldn't yet handle hundreds or thousands of cluster entities.

To prepare for our beta, we ran mental model research where we asked customers to draw their own mental map of their systems, and patterns started to emerge.

For the next iteration of the cluster navigator, I worked closely with our engineering team on the interface and with our data science team on adaptive ML thresholds for the cluster health metrics. We paired to figure out rules for how the map should scale, added an interactive table element as an alternative interface that could handle large-scale clusters, created a design system of accessible colors and patterns using blue, yellow, and red to represent health, and added filter and search functionality to make it easy to quickly find any pod or container.

I loved getting to work closely with engineers and data scientists on this project, and I still use a similar accessible color palette for traffic-light style health metrics today.

CA Technologies is now a Broadcom company.