Hey,
What is an Internal Developer Platform? What are the best practices for architecting them? What do you call the different layers/planes? What are golden paths? Time to become more concrete.
Thanks to McKinsey's platform teams (they are fantastic, btw), we now have reference architectures for AWS, GCP, and Azure ecosystems, with more coming soon. These standard patterns are based on real-world experiences, proven to work effectively for hundreds of setups. McKinsey is publishing articles and whitepapers on this soon. I got a sneak peek and wrote my own (cheesy, I know, but they authorized kindly).
They consist of 5 planes:
- Developer Control Plane
- Integration & Delivery Plane
- Resource Plane
- Observability Plane
- Security Plane
I wrote (extensive) whitepapers to explain the design choice and the interplay. And especially how they leverage the workload specification Score and Humanitec’s Platform Orchestrator.
You can get them here:
- Reference Architecture Whitepaper for AWS ecosystems leveraging Github, Backstage, Humanitec, Github Actions, Terraform, and more.
- Reference Architecture Whitepaper for GCP ecosystems leveraging Github, Compass Atlassian, Humanitec, Terraform, Vault, ArgoCD and more.
- Reference Architecture Whitepaper for Azure ecosystems leveraging Github, Backstage, Humanitec, Terraform, and more.
I know those are only whitepapers for now but good news: all those reference architectures are also available as packaged as code. As you know, I’m a big believer in Platform as Code for testability etc. McKinsey is in the process of open-sourcing the AWS one, and the GCP one is being built right now. Azure, Openshift and Multicloud are queued.
Stay tuned for my webinars on GCP and Azure reference architectures! In the meantime, check out the recording of my AWS webinar here.
There has NEVER been a better time to start building your platform. It’s getting more predictable and easier every single day.
Cheers,
Kaspar
