Hey everyone! I'm part of a small startup's platform team currently using AWS for our cloud needs, primarily relying on services like Aurora MySQL, EKS, and load balancers. We maintain site-to-site VPNs and direct connections, but those are mostly for our higher environments. We also manage our own Kafka queues with Strimzi running in EKS, and handle observability and SIEM solutions ourselves in that same cluster. Recently, we've been thinking about transitioning our lower test environments away from the cloud to save some costs – around $3-5k a month on AWS which we typically pass onto our customers. I'm reaching out to gather some thoughts on this: 1. If we go this route, should we consider using OpenStack to replicate our AWS environment for consistent testing? Or would that just complicate things? 2. Alternatively, could we just set up our own EKS cluster and MySQL to manage like we do on AWS? 3. Should we bypass bare-metal setups altogether in favor of cheaper clouds like DigitalOcean? 4. Is this whole switch worth it considering the effort involved? Thanks for any insights you can share!
4 Answers
Switching environments can introduce risk. If your test setup isn’t identical to production, you're not really testing effectively. Adding complexity for minimal savings isn't smart, especially with a small team. Assess your architecture and try to automate wherever you can instead!
You've got to weigh the benefits against the effort. Saving $4-5k a month sounds nice, but it's not a huge amount considering that even one junior engineer can cost around $5k monthly when you consider all the overhead. It could take you forever to recoup those costs. Maybe hold off for now and revisit later if cloud costs rise further!
That's a good point! We're definitely trying to keep things efficient with our limited resources.
Vendor lock-in can be a pain! Just remember to factor that in.
Nope, avoid this move entirely! Could you run different namespaces in a shared cluster instead? That might be a much more efficient way to manage costs while still ensuring your environments mirror each other as closely as possible. Adding a whole new stack only complicates things.
I'll definitely look into that! Thanks for the suggestion.
Honestly, I wouldn't recommend OpenStack unless you have someone experienced to manage it. Setting it up might not be too tough, but maintaining a lab environment like that can quickly get messy. Focus on something simpler like bare metal Kubernetes; it's easier to rebuild and maintain over time.
Have you had direct experience with OpenStack? I'm curious about how challenging it is to set up and manage.
Thanks for these points! They're super helpful for my discussions with management.