STPA for Software Systems–Illuminate the Unknown Unknowns
summary
Claim: STPA (systems theoretic process analysis) anticipates design flaws before production by modeling systems as control-feedback loops. Applied to a Google Maps road-disruption design, it finds seven design flaws — missed creates, race conditions, poor observability, onboarding and change gaps — and prevents the outage by changing the diff-based logic to read from the database. The exercise takes ~26 engineering hours and yields concrete loss scenarios and prioritized mitigations instead of vague risks.
AI-written summary · about
chapters (11)
- 0:00STPA introduction
- 1:22Traditional incident response limits
- 10:21What is STPA
- 15:19STPA four steps
- 22:29Control feedback loops
- 31:52Unsafe control actions
- 42:23Loss scenarios from UCAs
- 58:41Control path failures
- 1:05:25STPA for Google Maps
- 1:19:10STPA cost savings
- 1:22:22Q&A
related talks
01h 13mDesigning distributed systems with TLA+
22mFailing Over without Falling Over
40mSystems Thinking for Developers
07m"What does system design mean to tester?"
14mFeedback-guided Adaptive Testing of Distributed Systems Designs
50mDesigning Fault-Tolerant Software with Control System Transparency
27mArchitecting for Interconnected Global Ecosystems
41mDesigning software with complex configurations