Engineering Productivity Starts with Predictable Behaviour
summary
Argues that engineering productivity follows from measurable predictability: replacing cache‑heavy stacks with an SSD‑optimized NoSQL yields steady low latency rather than intermittent slowdowns. Shows benchmarks: 50k reads / 40k writes mixed workload that shifts latency up permanently on Cassandra, 1M transactions/sec vs 400k capacity for a like workload on Aerospike with fewer nodes, and real SLAs like 50 ms for LexisNexis/TIPS — built systems run predictably and the fixes actually work.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:49Unpredictable systems at Yahoo
- 3:59Engineering productivity needs predictability
- 6:23Hidden costs of unpredictability
- 9:56Unpredictable systems drain resources
- 15:10Benchmarks show unpredictability
- 18:00LexisNexis fraud detection
- 21:34TIPS instant payment system
- 23:19Predictability speeds development
- 24:22Conclusion
related talks
02h 31mDeveloper Productivity Engineering - Extended Session
45mDeveloper Productivity Engineering – The Next Big Thing in Software Development
29mEconomies of Speed
30mEngineering in Overdrive: How to Thrive When Tech Evolution Goes Full Gas
28mMeasuring Success
54mPerformance Architecture
26mBe Awesome By Being Boring
44mEngineering for Engineering's Sake