Types of Experiment Executions

Generally speaking, there are two ways we can run experiments. Or, to be more concrete, we’ll explore two even though are certainly other methods.

One-Shot experiments

We can run one-shot experiments by executing experiments on demand. We can say, “let’s run that experiment right now.” I’m calling them “one-shot” because such experiments would not be recurring. We’ll be able to choose when to execute them, and they will exist only while they’re running. We’ll be able to monitor the progress and observe what’s going on. We could run those experiments manually, or we could hook them into our continuous delivery. In the latter case, they would be part of our pipelines and, for example, executed whenever we deploy a new release.

Scheduled experiments

Later on, we’re going to explore how to create scheduled experiments that will run periodically. We might choose to do chaos every few minutes, every few hours, once a day, or whatever the schedule we’ll define is.

---

The next lesson will cover the execution of one-shot experiments.