From Gotehenburg Sweden. They’re ~90 employees, 45 engineers. They didn’t really have a dedicated platform engineering team at the start.
Each business area wants to spend the most time on business logic, but least time on:
- load balancing
- metrics
- security
- etc
They solved the “golden path” (see talk: Dan Bryant “from kubernetes past to what’s next” from kubecon valencia 2022)
- solve the common needs
- developers can focus on what’s important
uncomfortable insights:
- you can’t do “ALL the things”
- It’s valuable to NOT do something (e.g. you don’t have to keep maintaining it)
- upgrades are frequent and take time. (they spider out and sometimes have breaking changes)
- you probably don’t need a service mesh (lol; where “service mesh” is variable)
Approach: remove time consuming tasks and bottlenecks
- unblock the developers. They should always be able to self-serve their goals
- remove time consuming tasks from the platform teams (these are before devs b/c they are high leverage)
- remove time consuming tasks from the developers
They think it was an overall success. They wished they would have done more buy-in from the devs to better understand what good things they get.