How we leverage mechanisms at Amazon?
summary
Argues that building repeatable mechanisms beats ad hoc firefighting and good intentions, and shows exactly what was built: dashboards with P50/P90/P99, alerts, front-end telemetry (Sentry, New Relic), logging/metrics (Datadog, Splunk, Sumo Logic), feature flags (split.io), preprod regression runs and a CI/CD hook that performs auto rollback. Adds weekly operations reviews, launch readiness and metrics deep dives to drive adoption. Result: regressions surface earlier (caught hidden 10s page loads masked by averages) and incidents fall as rollbacks and monitoring stop customers being the primary detectors.
AI-written summary · about
chapters (7)
- 0:00Introduction and Background
- 1:29It's time to stop firefighting
- 2:42Sharing a Personal Story: Atlassian's Transformation
- 7:05Understanding Mechanisms
- 16:10Wrap up
- 18:14Mechanism Anti-Patterns
- 20:48Q&A Session
related talks
22mLeading software teams with context (visibility)
45mHow Leading Companies Are Scaling Their Security
44mThe Amazon Software Development Process
57mBuilding a Modern Security Engineering Organization
34mThe Road More Traveled: Putting Transformation Back on the Rails
41mAre We Getting Better Yet? Progress Toward Safer Operations
22mA Dual Approach to Accountability Engineering
12mRevolutionizing Software Development