Experiments

There are two kinds of experiments to choose from: Experiment and AsyncExperiment.

An Experiment is an iterable that updates its variables and routines in a synchronized fashion. On each iteration it spawns threads to update each routine, waits until all routines are updated, spawns threads to update each variable, waits until all variables are updated and then returns the state of the experiment. The benefit of this is that on each iteration, there is a well-defined state of the variables, and clear cause and effect with knobs being changed by routines and the ensuing response of the variables being recorded afterwards. The drawback is that the speed of iterations is limited by the slowest update process.

An AsyncExperiment is used the same way as an Experiment but the updating of variables and routines is executed asynchronously and as quickly as possible. The advantage is that iterations are generally faster and the state is updated more frequently. The downside is that iterations may miss some variable values and the causality between routines changing variables and the responses of variables to those changes may be less clear.