Anvil.server.timeout: What are my options

Anvil Q&A
1

What I’m trying to do:
I am building a front end for a simulation game (inspired by earth4all.life). The simulation model is large. One simulated year takes about a second and there are 120 years, so I am running into the server timeout on my free plan.

What are my options?

  1. Go to a paid plan and run the simulation as a background task.
  2. Can I run this “background task” on another server I control and uplink to it?
  3. Can I run the entire app on another server using the open-source anvil server?
  4. Something I did not think of?
    Eventually, i.e. for production, I wanted to use option 3 above. Is the timeout limit part of the anvil server, or only of the anvil server when run on anvil.works?
    Any guidance is much appreciated.
    Thanks, Ulrich
2

All of these are viable. We use option 2, queueing up our “simulation” jobs, reserving a distinct database table row to receive each result as it is computed.

Our compute engine is written in C++ for speed.

Option 4 might be to create a model engine in a faster (compiled?) language, wrap it in a Python package, and install that package in your server-side runtime environment. This isn’t always practical, though.

Edit: Numpy might be able to accelerate some kinds of calculations.

3

The timeout on background tasks is a feature of the anvil.works service, not inherently part of the Anvil server.

4

Would it be possible to restructure your code such that, instead of a single background task that performs simulations for all 120 years, you have a sequence of background tasks, each of which performs simulations for a smaller subset of years?

1 Like
5

Thanks for your ideas - chopping up the 120 years into smaller bits is probably the way to go, especially since all results are updated and available in a numpy matrix.

6

Thanks for the idea to write the compute engine in something faster than python (and then wrap it) Need to talk to our C++ folks!