Making Microservices work at Netflix
1 like602 views
Netflix engineers own the entire lifecycle of their microservices. They are provided tools to simplify development, deployment, operations, and support. Key aspects include standardized platforms, service discovery, logging, monitoring, configuration management, automated canary analysis, and fine-grained operational insights. The culture at Netflix emphasizes autonomy for engineers alongside principles of safety, velocity, and preparation for failure.
Making Microservices work at Netflix
- 1. The machinery around microservices Tools & practices for operating microservices @ Sangeeta Narayanan (@sangeetan) Director, Edge Developer Experience Dec 2016
- 4. >83M subscribers >1000 device types >42B streaming hours in 2015 600 hours of original content 1000s of playbacks per second 3 AWS regions; 3 AZs each
- 6. Subscribers/year (millions) 20 40 60 80 2007 2009 2011 2013 2015
- 7. Subscribers/year (millions) 20 40 60 80 2007 2009 2011 2013 2015 2016 ??
- 8. http://bit.ly/2gFd6mq Many things have to come together for a successful launch of this complexity
- 12. Making microservices work Tools, Practices, Culture http://bit.ly/2h7Q2NY
- 13. CDN data plane not shown
- 14. Architectural benefits ❖ Scalability ❖ Resiliency ❖ Velocity ❖ Autonomy
- 15. ❖ Dynamic Routing ❖ Traffic shaping ❖ Insights, debugging ❖ Traffic failover Custom cloud gateway
- 16. Philosophies ❖ Own your service - operate what you build ❖ Move fast - optimize for velocity ❖ Stay safe - prepare for failure ❖ Focus on automation http://bit.ly/2fydxy6
- 17. Develop Deploy Operate Support App Development & Delivery Insights Netflix engineers own the entire lifecycle of their services. They are provided a rich set of tools and libraries to simplify their tasks.
- 18. Development ❖ Standardized platform ❖ IPC ❖ Service discovery ❖ Logging, Monitoring ❖ Configuration Management ❖ and more ❖ DevEx tooling ❖ API discovery and documentation ❖ Usage tracking and lifecycle management ❖ Error simulations and behavior overrides
- 21. Deployment
- 22. ❖ Automated canary analysis ❖ Self-service canaries for developers ❖ Automated regional deployments ❖ Choice of deployment strategy ❖ Pipelines for config changes
- 27. Support ❖ Diverse audience ❖ Internal partners: service consumers, dependencies, customer services ❖ External partners ❖ Production and lower environments ❖ Micro level insights
- 28. Fine grained insights via real time event stream queries
- 31. Operational Insights ❖ Aggregate health dashboards ❖ Real time alerting ❖ Anomaly detection ❖ Automated correlation analysis
- 39. Failure testing Drills and “game-day” exercises Latency and error simulations
- 42. Culture
- 47. http://www.martinfowler.com/articles/microservices.html People Practices Technology
- 49. Thank You