Need for SPEED: Site Performance Efficiency, Evaluation and Decision
summary
Shows that common CPU utilization numbers lie: an Intel two‑socket 16‑core system with hyperthreading yields 64 logical CPUs, and running SPECjbb2005 pinned to 32 versus 64 threads produces roughly the same throughput while reported utilization moves from ~50% to ~100%. Proposes RUE (resource usage effectiveness) = resource used per work done and a speedup ratio, and demonstrates a Simpson’s paradox where overall speedup is 1.09 but three subgroups show 0.87, 0.99 and 0.84. Provides code pointers on GitHub to reproduce the analyses.
AI-written summary · about
chapters (8)
- 0:00Introduction and scope
- 1:32Cpu utilization metrics
- 4:33Hyper threading impact
- 7:56Throughput experiments
- 11:30Resource usage effectiveness
- 15:15Evaluation at scale
- 20:24Understanding Simpson's paradox
- 23:49Vector analysis and summary
related talks
35mCPU Utilization: The Hidden Cost of Running Hot
01h 06mPerformance engineering: being friendly to your hardware
47mPractical Performance Analysis
38mPractical Performance Analysis
40mPerformance Testing Done Right
58mNeed for speed 8, performance tuning of your web application
01h 05mPerformance Optimization in Software Development - Being Friendly to Your Hardware
15mOptimizing Application Performance on Kubernetes