Shiny in production: Principles, practices, and tools
summary
Argues Shiny is production‑ready and demonstrates concrete tooling and workflow: use shinyloadtest to record realistic traffic, use profvis to find hotspots, move heavy ETL out of the server and precompute feather files, then apply renderCachedPlot‑style plot caching. Example app targets 100 concurrent users on a 16‑core machine; median session latency falls from ~210s to ~50s after ETL and to ~20s with plot caching. Caches can live on disk, shared filesystem or Redis; approach noticeably improves scalability and predictability.
AI-written summary · about
chapters (13)
- 0:00Intro
- 1:56CRAN Wales
- 5:39What is production
- 10:25Can Shiny be used for production
- 13:46Organizational challenges
- 15:38Technical challenges
- 23:04Run your shiny app
- 23:54Shiny Load Test
- 26:02Profiler
- 29:27ETL
- 33:41Plot caching
- 39:31Summary
- 43:43Best practices
related talks
20mWhat I learned from running Shiny in production
35mBuild Simple and Scalable Apps with Shiny
15mA Robust Framework for Automated Shiny App Testing
44mShiny: Data-centric web applications in Python
58mThe Past and Future of Shiny
01h 32mCreating Production-Ready Shiny Apps
14mR Shiny - From Conception to the Cloud
25mUnderstanding reactive execution in Shiny