A proxy should never retry. This blog post explains why. This sim doesn't actually consider the possibility of a failure in the backend. Instead it is trying to highlight why it should always be the client and not the proxy that handles the failure by demonstrating how bad the performance implications of a retrying proxy are.
You can download the source from github and open it in the browser via file:// URL or run from your own website
Parameter description:
- tmin - the minimum time at which your backend is likely to respond
- tmax - the maximum time at which your backend is likely to respond
- tret - the time at which the proxy retries
Legend:
- blue - the probability that request will be fulfilled at this time
- green - the probability that request will be fulfilled by this time
- red - the probability that request will be fulfilled by this time with retries