Entrevistamos a David Lluna
summary
Defines "controlled" as isolating a single variable, fixed hypothesis and preplanned metrics, and enforces it with feature flags and canary deployments. Argues for concrete safeguards: sample-size calculations, automatic rollbacks, and a safety gate that cuts traffic at a preset error threshold. Claims benefits in reduced blast radius and faster learning, quantifies risk by showing acceptable loss as a percent of traffic, and cites using real-time metrics and A/B stats to decide success.
AI-written summary · about
related talks
27mDEVOPS DAYS MTL 2018 - Lenore Adam
21mTesting in Production
01h 15mTaking Control Of Your Test Environment
22mTesting in Production
56mControlling Quality in Software Development
25mGo Faster, Be Safer
34mHypothesis to Impact: Mastering the Art and Science of Controlled Experiments
25mPushing Code: Don't Forget to Flag Your Canaries?