How to Build a Smart Profiler for Rails
summary
Skylight claims to deliver honest response times by storing histograms and 95th percentile rollups, and it actually handles the scale of real traffic — the system processes on the order of one billion requests per month and uses protobuf, Kafka, Storm and Cassandra to batch and roll up data. A low‑overhead Rust agent plus a coordinator keeps memory down, a sampling profiler taps Ruby 2.1 stacks, and an Ember+D3 frontend gives realtime aggregate traces and automated inspections like n+1 detection so the product actually works end to end.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:15What is Tilde
- 7:55Problems with Tilde
- 9:10Performance Monitoring
- 11:36Memory Management
- 13:29Coordinator Object
- 15:44Kafka
- 25:03Inspections
- 26:59Feedback Loop
- 29:24Skylight Demo
- 33:00Stack Sampling
related talks
39mProfiling and Benchmarking 101
37mUsing Skylight to Solve Real-World Performance Problems
37mThe Future of Ruby Performance Tooling
26mRails Performance Monitoring 101: A Primer for Developers
41mScalable Observability for Rails Applications
21mBuilding a Performance Analytics Tool with ActiveSupport
32mRails Performance First Aid
30mRails Performance monitoring: A primer for developers