General

·

·

8 min read

"We could build this ourselves." Yes. Here's what that actually means.

Tony Como

COO & Co-Founder, Juno Innovations

Engineering workbench with tools and a whiteboard

ON THIS PAGE

“We could build this ourselves.”

We hear it on a lot of calls, and it's usually true. Kubernetes is open. Helm is open. Most of the building blocks are open. A capable team can stand up something that launches workloads for its users.

So when someone says it, I don't argue. I ask them to tell me what they think we do.

The answer is usually a deployment UI and a set of Helm charts. That's part of it. It's the part you can see in a demo.

What you're signing up for

In broad strokes, a platform like this has to know who every user is, put their work on the right hardware, attach the right storage with the right permissions, and keep each group inside its limits. It has to do that the same way every time, for people who will never open a terminal. Then it has to keep doing it through every Kubernetes release, every driver update and every new machine someone plugs in.

I'm not going to walk through how we handle each of those. The details are the work, and every environment we've been in handles them a little differently. Each one took us longer than we expected, and this is all we do.

The part nobody budgets for

The build is the easy number to estimate. Maintenance is the one that sneaks up on you.

Kubernetes ships about three minor releases a year. Storage gets upgraded. New GPUs show up and the templates need updating. Dependencies pick up security fixes. None of that stops after launch day.

When the person who built it needs to move on

A university research computing group we spoke with recently is a good example. One of their engineers built a GPU-sharing setup for thousands of students, and it works. They built tools most people wouldn't attempt. They also know they're the only one who can run them. In their words, that's “a really, really dangerous spot to be.”

Their leadership wants them free for other projects. That's the real cost of a homegrown platform, and it rarely shows up in the build estimate. The person who built it can't stop maintaining it.

We didn't pitch them on throwing that work away. We talked about running Orion next to it, taking on the parts nobody should have to babysit and keeping the parts that are theirs. By the end of the call, they wanted to try it.

“We can pretty much do all this”

That pattern keeps repeating. An HPC team at a large public university started a call with us saying they could pretty much do all of this themselves. By the end of it, they were talking about putting Orion on their test cluster as a proof of value.

In both cases the shift came during the demo, once they saw it running. Nobody had to be talked out of building. They just got to see what they'd be signing up to maintain.

When building it yourself is the right answer

Sometimes it is, and I'm not going to pretend I know which side your team is on. Every team is different.

Your requirements are narrow and stable. If you need exactly one workload type, in one environment, on one storage backend, and that won't change, a general-purpose platform is more than you need.

Your team has the capacity, and not just the capability. Those are different things. A capable team with no free cycles ends up with a half-maintained platform.

The available tools don't fit your constraints. Some environments have requirements no platform handles, and building in-house may be the only option.

The available tools cost more over three years than building and maintaining your own. Then build your own.

What you'd be starting from

If you go the other way, Orion runs in your environment, on your hardware or your cloud account. The Orion Apps store has over 50 plugins, and any Helm chart can be packaged as one. Your workloads stay standard Kubernetes, so if Orion went away tomorrow, your Helm charts would still work.

The teams that decide well do the actual math. They work out what it costs to build, maintain and evolve over three years, who has to own it, and put that next to the alternatives.

Your team can build it. The question is whether the business wants to run a software shop on the side, and who gets the call at three in the morning when it breaks.

If you want to run that math for your environment, we'll do it with you. Book a time with our team

WRITTEN BY

Tony Como

Tony runs go-to-market at Juno, from first outreach to production deployment. He works across sales, partnerships, and operations to help technical teams move faster without adding infrastructure overhead.