Reactive vs Imperative

This is a demo project containing various modules:

• slow-server server with a slow response (default: 5 seconds, increase to have more time to illustrate threading behavior), starts on port 9000.
• reactive-app reactive application accessing slow-server, starts on port 9100. Limited to 16 threads.
• servlet-app servlet-stack application accessing slow-server, starts on port 9200. Limited to 16 threads.
• load-generator generates load (100 requests) to reactive-app/servlet-app.

Both applications (reactive-app, servlet-app) dump their thread count on the console.

Load Generator on reactive app

Running the load generator should let you see something like:

2017-11-06 11:42:15,159  INFO          load.GenerateLoadForReactiveApp:  40 - Starting requests...
2017-11-06 11:42:15,340  INFO          load.GenerateLoadForReactiveApp:  46 - Running...
2017-11-06 11:42:21,104  INFO          load.GenerateLoadForReactiveApp:  52 - Done!

Load Generator on servlet app

Running the load generator should let you see something like:

2017-11-06 11:42:19,674  INFO           load.GenerateLoadForServletApp:  40 - Starting requests...
2017-11-06 11:42:19,863  INFO           load.GenerateLoadForServletApp:  46 - Running...
2017-11-06 11:42:55,208  INFO           load.GenerateLoadForServletApp:  52 - Done!

If you pay notice to the completion time, you'll notice the difference in completion time. The end-to-end reactive approach takes just about the duration of the wait while the imperative (Thread blocking) takes roughly 31 seconds (100 requests / 16 threads = concurrency factor 6.25 * 5 ~ 31.25 sec).