Adrian’s Greatest Hits, B-Sides & Re-issues
summary
Reruns of Netflix architecture show concrete moves: migrate to AWS, replace Oracle with Cassandra (three replicas per write), run Chaos Monkey to force stateless services, and shift deploys to per-team microservices with library-based interfaces and faceted IDs so teams release independently. Practical failures and wins include surviving an AZ outage while others failed, telescoped timeouts and retry policies to prevent retry storms, a Go-based simulator for huge service graphs, and serverless prototypes that ship features in days.
AI-written summary · about
chapters (8)
- 0:00Start
- 6:36Untangled Service Interfaces
- 12:29High Availability
- 14:18Chaos Monkey
- 22:35Microservices version
- 26:51Non-Destructive Production Updates
- 28:45Speeding Up The Platform
- 30:41Timeouts and Retries
related talks
42mPlanning To Fail
41mDystopia as a Service
39mResilience & Chaos Engineering in the Cloud
35mBuilding Systems that Last
51mApplication Stacks for the Modern World
31mDevOps in the Cloud: Case Studies of Amazon.com teams and their resilient architectures
45m⚡A Resilient Application in a Partially Degraded World ⚡
20mResilient and well-architected apps with chaos