Explore a comprehensive conference talk on defining reliability in platform engineering, focusing on Google's Serverless SRE team's approach to detecting and measuring latency regressions. Dive into topics such as total latency distribution, request delivery latency, and the limitations of SLOs. Learn about the 2-Sigma Technique, overload scores, and impact analysis for tracking platform performance. Discover practical applications including streamlined diagnosis and approximate cohort A/B testing. Gain valuable insights into customer-centric performance measurement and statistical approaches to platform reliability from an experienced SRE at Google Cloud.
Why Is My App Slow - Defining Reliability in Platform Engineering
Overview
Syllabus
Intro
Serverless platform is amazing
"My app is slow"
The platform is slow
Total end-to-end latency distribution
Request delivery latency
Goal
Reliability in practice
Applying to the model
Stationarity
2-Sigma Technique
Mechanics
Overload score
Impact analysis
FAQ
Backtesting
Limitations
Other applications
Streamlined diagnosis
Approximate cohort A/B testing
Conclusions
Outro
Taught by
GOTO Conferences
