The Universal Laws of Performance
summary
The talk explains universal performance laws from queuing theory, notably Gunther’s Universal Scalability Law, to model and predict application performance without exhaustive load testing. It shows how to apply these laws to practical problems such as capacity modeling and load estimation, focusing on intuition and concrete application rather than detailed math.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:31Parallelizing for Speed
- 3:30Limits of Parallel Acceleration
- 4:37Example: Parallel Collections
- 10:15Anatomy of a Queue
- 12:28Measurements of a Queue
- 15:50System Modeling
- 20:33Measuring API Latency
- 23:28Behavior of a Closed System
- 25:43The Universal Scalability Law
- 34:33Application Examples
- 40:45Modeling Systems
- 42:07Anticipating Failure Modes
- 43:18To Go Further...
related talks
38mA Practical Look at Performance Theory
30mScalability Is Quantifiable: The Universal Scalability Law
46mQueueing Theory in Practice: Performance Modeling for the Working Engineer
43mThe Math of Scalability
40mSystem Performance and Queuing Theory - Concepts and Application
30mQueueing Theory, In Practice: Performance Modelling in Cloud-Native Territory [I]
06mPractical performance theory
35mSystems Scale Summer Q&A Week 1