Operation Molnett - intro to building a Cloud Provider from "scratch"

I've worked with Cloud Infrastructure for as long as I can remember, dating back to the Hostgator / cPanel days. Now that I think about it, that's actually exactly half of my life! nuts. Feels like it has been forever.

My passion for Infrastructure has led me to work on absolutely amazing projects, which includes Embark Studios, Candy Crush, Google Partner, and I've met some people that I now consider my closest friends, co-founders and even my life partner!

Despite all the amazing experiences, I've had a constant feeling that everything was too complicated. Consensus algorithms and other kinds of distributed programming adds a large amount of cognitive load to operators, and it just gets worse the bigger you get.

This, and many other things, culminated in me moving forward from my job at Embark Studios and head towards the wild territory of cloud computing to create something that fits my personal requirements and mental model.

Inspired and vigorous

The past years Fly.io has written countless posts about the methods they use to solve this problem. They work specifically with making applications more easily deployed at the edge, which is quite niche. Fly has written about their reasons of initially running with Nomad instead of Kubernetes, creating their own use-case specific Load Balancer rather than shoehorning an existing technology, and about tons of other topics.
It has been a joy to read their thorough posts on how they deploy IPv6+Wireguard as an overlay network, their view on modern orchestrators and many more. And it led me to thinking of how I'd do it, if I had all the freedom in the world to do it.

And here we are! 😊

From an outside perspective, it might seem brave to take on such a huge problem space without being backed by major company. On the contrary, I feel that it gives me an edge!

On a personal level, being able to fully focus on a specific piece of tech without the requirements of a formal routine (stand ups, meetings, etc.) has proved to be both healthy and intellectually rewarding. Just a few days ago I spent an entire evening and night (studying at 2AM is the best. fight me) reading the Google Omega paper and implementing a toy orchestrator in Rust. I am not so sure that would be possible at a standard job without the no-sleep hangover.

And equally as important, I am not afraid to step on anyone's toes or waste someone's time. If I'd like to hack together a PoC of running a Wireguard overlay network on top of Firecracker VMs, I can just go at it and see where it leads. No one can tell me otherwise, and so far it has been very productive!

I've broadened my horizons on what modern infrastructure could look like if we just spent a little more time making specific solutions, rather than resorting to running all our workloads on top of the gigantic and generic Kubernetes, pulling our hairs out when it inevitably comes back to bite us unless we take fine care of our clusters.

What that means in practice, I'll get back to you on that in the next publication!

Operation: Molnett

Out of all these experiments, a tangible entity has spawned! Molnett is the brand and company that will be running the platform that these years of work will accumulate to. I'm very excited to show you all what I've been working on later this fall. Until then, make sure to subscribe to the newsletter for future updates and to follow our Github for when I inevitably publish some of the work.

Hear from you soon 💜
Jonathan Grahl