Windows at Scale
summary
Just Eat migrates a Windows platform into AWS, standardizes on Packer-built AMIs, CloudFormation stacks and autoscaling groups, and moves from ~10–20 yearly changes to ~1,000 deployments by 2014. Monitoring uses statsd/Graphite and ELK with structured JSON logs; alerts live in source control. Deploy strategy doubles an ASG, waits, then halves it; runtime updates cut deploy time from 20 minutes to 2 minutes. Platform has ~120 services, feature flags for releases, and TeamCity agents spun up from images on demand.
AI-written summary · about
chapters (26)
- 0:00Introduction
- 0:38Who am I
- 2:18Questions
- 2:38Windows at Scale
- 5:00Just Eat
- 5:34Scale
- 16:23Shipping to AWS
- 19:36Elastic Search
- 22:18Alerts
- 24:17Structured logs
- 26:54Deployment
- 29:02Remote execution
- 31:48Octopus Deploy
- 34:33Deploying dependencies
- 37:19Deploying images
- 39:11Windows support
- 41:19Database
- 43:08Clusters
- 44:43Release cycles
- 47:37Environments
- 49:36Pipeline
- 51:42Continuous Testing
- 53:37Writing Contracts
- 55:10Monitoring
- 58:39Cluster scheduling
- 59:19Lambda
related talks
25mCommon Platform: Now with Added Windows
22mFireside Chat with AWS
23mA Journey to Automated Infrastructure
22mMigrating Windows workloads to AWS
36mPractical Cloud Native: What Works, What Doesn't
33mNavigating the Seas of Transformation: Migrating from EC2 to Kubernetes
47mDeploying Like a Boss: From 2 Servers to 20,000
24mSleeping soundly even when scaling your app to 200k request/seconds