Salmon v0.2.3 can handle up to at least 100 simultaneous users interacting with the system and answering questions. Here’s the setup:
Launched Salmon on a Amazon EC2
Simulated client response time using the most similar task I could find 2. Response times were a modified Gaussian random variable with mean 750ms and standard deviation 250ms (modified so always greater than 200ms).
Each client for 50 responses.
Here are the performance results with that setup:
This performance is pretty good! Most of the responses return in around 400ms. How do the extremes perform?
45 of a users queries will return in less than 500ms if the user answers 50 questions and there are 102 simultaneous clients.
Specifically, a spatial configuration task with 3 elements. I pulled number from Figure 6 of Palmer et. al [palmer].