How Close is Ruby 3x3 For Production Web Apps?
summary
Benchmark runs a seeded, repeatable user workload against Discourse forum on a single dedicated r4.2xlarge EC2 VM using Puma 10 processes 60 threads and measures percentiles, time-to-first-request and warm-up effects. Results show Ruby 2.4.1 ≈150% of Ruby 2.0 throughput and about 30% faster time-to-first-request, bootsnap adds ~15% boot speed, and 5,000 warm-up iterations yield ~5–7% higher throughput. Benchmark code, JSON outputs and Packer-built images are available to reproduce or vary the configuration.
AI-written summary · about
chapters (25)
- 0:00Introduction
- 0:51RealWorld Benchmarks
- 2:33Discourse
- 3:51Random Requests
- 4:57Example
- 5:29Configuration
- 8:18Warmups
- 10:07RealWorld Benchmark
- 11:17Graphs
- 12:22Discourse Version 15
- 12:43Discourse Version 18
- 13:25Summary
- 13:45Does Ruby need warmup iterations
- 15:05More warmups equals better throughput
- 15:42Slope
- 15:52Warmup iterations
- 16:09Generational garbage collectors
- 17:35Discourse and BootSnap
- 19:15Time to first request
- 19:55Conference WiFi
- 21:41JSON Files
- 22:24The Code
- 24:34Future Work
- 26:48Conclusion
- 28:48Questions
