Week 1 Live Q&A
summary
Live Q&A covers concrete ops choices and measured outcomes: distributed tracing is valuable once services interact widely and tracing overhead is quantified via linear regression to estimate trace-start overhead plus dummy services for SDK continuation cost; codegen compilation uses a global cache to raise hit rates from 99.5% to 99.95%; streaming pipelines run now on Pub/Sub and cloud streaming APIs and cost more than batch while Kafka remains pluggable; capacity is set statically by owners and Raz handles reassignment with penalties amid hundreds of reservations per day.
AI-written summary · about
chapters (18)
- 0:00Live Q&A introduction
- 0:45When tracing becomes valuable
- 2:31Stream processing trade-offs
- 3:56Capacity request management
- 5:42Linear scaling and CPU efficiency
- 6:48Deciding trace data contents
- 8:11SAT solvers for capacity
- 9:15Event pipeline datasets
- 11:05Telemetry tools overview
- 12:51Bias in trace data
- 15:38Server reassignment
- 17:24Streaming vs batch costs
- 18:27Query compilation overhead
- 20:14User interaction with tracing
- 21:16Failure domains in maintenance
- 22:38Pub/Sub vs alternatives
- 24:06Schema recommendation confidence
- 25:09Closing remarks
related talks
13mSystems @Scale Summer Live Q&A Session #3
17mDeveloper Experience Q&A
21mLive Q&A Session #2 – Adi Gangadi
20mSystems @Scale Summer Live Q&A Session #2
27mHow to be data-driven when you aren’t Netflix (or even if you are)
26mProfiling Mastery: Decode Your Data and Elevate Observability
24mWhy Does Continuous Profiling Matter to Developers?
45mIntroduction to Akka Stream