The Amazon Software Development Process
summary
Amazon organizes development as hundreds of autonomous two pizza teams each owning a service end to end, uses an internal release tool called Apollo and a CI/CD system called Pipelines, and records millions of deployments per year. Releases use pessimistic deployments that roll changes to one host then one AZ then region to limit blast radius. Operations enforce accountability via weekly “wheel of fortune” reviews and business metrics, and product decisions start with a working backwards PR/FAQ and press‑release style spec.
AI-written summary · about
chapters (10)
- 0:00Introduction and Amazon overview
- 3:20From monolith to microservices
- 6:50Two pizza teams and autonomy
- 9:46Self-service tools and pipelines
- 17:08Pessimistic deployments
- 22:30Automated testing and shift left
- 26:26Obsession with operations
- 33:03Business metrics and accountability
- 36:16Working backwards from customers
- 41:54Trade-offs and closing remarks
related talks
44mThe Amazon Software Development Process
21mDevOps Culture at Amazon
50m Drive change with Amazon’s culture of innovation
27mDevOps in the Cloud
31mDevOps in the Cloud: Case Studies of Amazon.com teams and their resilient architectures
10mInterview
41mTowards Operational Excellence
38mAccelerating innovation through AWS Cloud technology