Optimizations in Multiple Dimensions
summary
Starts with concrete claims: JIT warmup makes later runs faster, MRI allocates 16KB pages for objects, small hashes use a flat array until 3 keys, and a CPU-bound job runs 3s dev vs 20min prod under realistic load. Caching costs include network latency (example 10ms) and eviction strategy; aim for cache hit rate >99% (drop cache if <90%). Measure latency and throughput separately, record metrics with Librato or New Relic, then tune and remeasure.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 0:45What is optimization
- 2:46Metrics
- 5:28Performance
- 13:39Communication
- 14:08Tradeoffs
- 14:41CPU vs RAM
- 17:43Small vs Large Data Sets
- 18:19Most Data Sets are Small
- 19:50MRI
- 21:42Caching
- 23:57Cache Invalidation
- 26:19Cache Hit Rate
- 28:14Readability
- 30:27Performance in Production
- 31:48Summary
- 33:45Outro
related talks
34mCache = Cash! 2.0
50mDesigning for Performance
29mOptimizing Production Performance with MRI JIT
37mCache is King: Get the Most Bang for Your Buck From Ruby
36mMake It Fast
51mDesigning for Performance
46mThe Impact of Micro-Optimizations on Large-Scale Apps
40mLudicrous Speed: Designing For Performance on the JVM