What drove us to build Control Plane?
We didn’t have a choice. We looked around and realized this had to be done.
We helped lots of people use K8s and knew it is no fun.
Those of us tasked with delivering microservices know that DevOps involves spending half of our time on Ops. When I thought about this fact, it bothered me. After all, does Pepsi need more availability than Coke? Do they need more security or less latency? If the answer is “no,” why do they do “Ops” in entirely different ways? Sure, they both use open source tools like Prometheus, Grafana, Istio, Kubernetes, and many more – but they do so in utterly different ways. If their desired outcomes are the same – why do they need to “re-invent” how to achieve common goals such as availability, observability, low-latency, least-privilege access controls, auditability, etc.? Do they need to operate all these components? There must be a better way. In short – they don’t need to. I had an “aha” moment. Imagine if every company invented their very own programming language. Go isn’t good enough. Company A needs GoFoo, and company B needs GoBar. If each company developed its programming language or database management system, it would be absurd. It would work but be wasteful, and we know that intuitively. Our intuition should also tell us that stitching together an Ops complex apparatus is as wasteful as inventing a different language in each organization. This realization led me to assemble the most incredible software architects, ops gurus, SREs, and developers I know. I was fortunate to rub shoulders with exceptional engineers at VMWare, SAP, and my old startup BiTKOO. I sold them on the idea of codifying the Ops part “as a service”. We flew off to a remote area and closed ourselves in a room. We decided to envision what was possible. In a future post, I’ll describe that initial and crucial week. We dreamed big, but we stuck to our dream, and I could not be any prouder.