Thoughts on rolling your own infrastructure versus managed infrastructure 💻.
Вставка
- Опубліковано 24 сер 2024
- A co-worker reached out and asked why I was suggesting managed infrastructure as a first option versus rolling our own. I based my reasoning primarily on the history and size of the team which was followed by cost analysis. YMMV
There are some things that you don't want to manage based on your team size e.g. bootstrapping Kubernetes (although, k3s and other solutions do make this easier), maintaining your own selenium grid, etc.
Personally, I feel like for most services that aren't core/critical to an application/workflow/whatever, the only real question for managed services vs rolling your own is the relative cost - If you roll your own infra, that comes at a cost to the org, and typically not a cheap one. You've got to pay for x engineering hours each week to maintain the service at a minimum, and generally speaking if you're rolling your own infra, you'll want to be actively improving it as well. My org is looking into MSK with AWS for exactly this reason - we spend a bunch of engineering hours working on bringing our clusters up to a place where we want them, while also improving existing workflows, all of which is coming at a cost of other projects that could have a much bigger impact on the business.
Really, I think if you're rolling your own infra, you need to have some pretty strong reasons to do so (e.g. security, customisations, etc). A lot of teams will claim to have these reasons, but all too often they're not accurately estimating the costs/benefits or have strong Not Invented Here problems.