shinyloadtest
Tools for load testing Shiny applications
The shinyloadtest package, along with the shinycannon command-line tool, enables load testing of deployed Shiny applications to estimate how many concurrent users an application can support. It helps developers and administrators determine if their app can handle expected traffic levels.
Load testing with shinyloadtest identifies performance bottlenecks and guides infrastructure, configuration, or code improvements. The workflow involves recording a typical user session, replaying it in parallel to simulate multiple simultaneous users, and analyzing the results through automated reports. This approach provides concrete evidence that Shiny applications can scale to handle large numbers of users when properly configured.
Contributors#
Resources featuring shinyloadtest#
Veerle van Leemput | Analytic Health | Optimizing Shiny for enterprise-grade apps
Can you use Shiny in production? A: Yes, you definitely can.
Link to slides: https://github.com/RStudioEnterpriseMeetup/Presentations/blob/main/VeerlevanLeemput-OptimizingShiny-20220525.pdf
Packages mentioned: ⬢ shiny: https://shiny.rstudio.com/ ⬢ pins: https://pins.rstudio.com/ ⬢ plumber: https://www.rplumber.io/ ⬢ blastula: https://github.com/rstudio/blastula ⬢ callR: https://github.com/r-lib/callr ⬢ shinyloadtest: https://rstudio.github.io/shinyloadtest/ ⬢ shinycannon: https://github.com/rstudio/shinycannon ⬢ shinytest2: https://rstudio.github.io/shinytest2/ ⬢ feather: https://github.com/wesm/feather ⬢ shinipsum: https://github.com/ThinkR-open/shinipsum ⬢ bs4Dash: https://rinterface.github.io/bs4Dash/index.html
Timestamps: 2:44 - Start of presentation 5:41 - What qualifies as an enterprise-grade app? 10:46 - UI first / user experience / prototyping 13:20 - Separating code into separate scripts and creating code that’s easy to test 17:15 - Golem 19:28 - Functionize your code 20:50 - Rhino package, framework for developing enterprise-grade apps at speed 22:33 - Infrastructure, how do you bring this to your users? (lots of ways to do this. They do this with R, pins, plumber, rmd, blastula, and Posit Connect on Azure) 31:17 - Optimizing Shiny (process configuration, cache, callR, API, feather) 47:35 - Testing your app (shinyloadtest and shinycannon) 50:23 - Testing for outcomes (shinytest2) 52:15 - Monitor app performance & usage (blastula, shinycannon, usage metrics with Shiny app)
Questions: 57:38 - What’s the benefit of using pins rather than pulling the data from your database? 59:30 - Are there package license considerations you had to think about when monetizing shiny applications? 1:00:45 - Do you use promises to scale the application? (they use CallR) 1:01:49 - For beginners, golem or rhino? 1:02:50 - The myth is that only Python can be used for production apps, what made you choose to use R? 1:05:12 - Is feather strictly better than using JSON? 1:06:38 - Where do you see the line between BI (business intelligence) and Shiny for your applications? 1:08:36 - Any tips for enterprise-grade UI development? Making beautiful apps (bs4Dash app) 1:10:25 - Have you found an upper limit for users? 1:12:19 - Any tips for more dynamic data? (optimizing database helps here) 1:13:50 - Where do you install shinycannon? (on our development Linux server) 1:15:00 - Can you share other resources or examples of code? (Slides here with resources: https://github.com/RStudioEnterpriseMeetup/Presentations/blob/main/VeerlevanLeemput-OptimizingShiny-20220525.pdf )
For upcoming events: rstd.io/community-events-calendar Info on Posit Connect: https://www.rstudio.com/products/connect/ To chat with Posit: rstd.io/chat-with-rstudio