Knowledge is Imperfect - Acting on Stale, Inconsistent or Missing Data
summary
Engineered command and control and telecom systems that survive flaky links by using asynchronous replication, optimistic concurrency and manual fallbacks; deployed a replicated DB over a 19.2 kbps multiplex link that prioritizes Priority‑One items during exercises with ~50,000 soldiers. Built a VoIP/voice‑over‑ATM platform of 2 million lines of Erlang that guarantees service emergency calls at 500% load. Uses UDP heartbeats to detect split‑brain, models acceptable knock‑out units (~2,000 subscribers) and treats specs as code when they fail.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 0:24Outline
- 1:49Alaskan Adventure
- 3:20Ericsson adventure
- 6:39C2: Distinctive Challenges
- 15:46Telecom: Special Challenges
- 16:52Device Management Challenges
- 18:06Decision Support Basics
- 22:47The What
- 23:39Abstractions
- 25:06Ulf's Law of Information Management
- 28:44The 'What' for Developers
- 31:42Dealing with requirements
- 35:00Specifications
- 37:34Trust/verify
- 40:24Modeling data lifetimes
- 45:31Diagnosing absence of signal
- 47:32Knock-out Units
- 50:46Conclusion
- 52:41Questions?
related talks
45mChasing unicorns : 0 dataloss and 99.999% availability
41mDesign (Failures) Case Study
01h 19mBuilding an Efficient Replication System for Thousands of Terminals
55m12 Calm Years of PostgreSQL in Critical Messaging
23mExperimental approach towards Replication in Postgres
40mLatency, Throughput & Fault Tolerance: Building a Modern Database
29mThings might go wrong in a data-intensive application
44mBeauty and the Beast: your application and distributed systems