Bootstrapping in Data Analysis

How would you define the concept of "bootstrapping" in the context of data analysis? Can you provide an example scenario where it would be beneficial to derive smaller samples from a larger dataset using bootstrapping? Discuss the pros and cons associated with this technique.