comses / miracle

Repeatable data analysis workflows for computational models
1 stars 3 forks source link

Stress test the demo server #75

Closed warmdev closed 7 years ago

warmdev commented 8 years ago

This is important if we are going to serve 30 or more concurrent users. Mainly,

alee commented 8 years ago

Can use http://locust.io/ or apache ab to do something simple - agreed, we should figure out the max number of supportable users N. We may need to structure the workshop to have < N participants looking at the app at a time..?

alee commented 7 years ago

radiant becomes a clear bottleneck with over 2-3 concurrent users, and I'm guessing DeployR would also be an issue there. need to revisit that part of the architecture

warmdev commented 7 years ago

The problem with Radiant is that Shiny Server open source edition is single-threaded, so that can only be solved by one container per user. For DeployR, we can do one or more containers per project based on load.

BTW Docker 1.12 introduced a swarm mode which makes scaling a lot easier, and I assume it will come to dockerpy soon.