MathOps or the Mathematics of Percentiles in Monitoring
summary
Explains using MathOps for percentile calculations in monitoring systems, focusing on algorithms and techniques for estimating percentiles efficiently. Covers practical tools and implementations to compute and track percentile metrics in production monitoring.
AI-written summary · about
chapters (16)
- 0:00Monitoring problem
- 1:23Queuing theory
- 3:29Time series charts
- 5:15Distribution density limits
- 7:45Mean unsuitable
- 10:54Outliers and mean instability
- 14:47Normal behavior goal
- 17:16Percentiles in systems
- 20:45Percentile-only pitfalls
- 24:1399th percentile instability
- 27:06Forgotten sessions impact
- 28:53Simple monitoring traps
- 30:18Histograms and metrics
- 33:07Data aggregation limits
- 35:11No universal rules
- 36:13Alert goals
related talks
52mMonitoring Design Principles
38mKnow your data: The stats behind your alerts
36mStatistical Methods and Tools for System Performance Analysis
32mSoftware Performance: A Shape Not a Number
06mPerformance is about more than time series charts
16mShrinking the Observability Bill
58mSystem Metrics: Collecting the Gotchas
38mThe Metric System: Making Correct Performance Measurements