The algorithm is simple. Assume you have a lot of things to choose from, and it’s expensive to test a choice to see how good it is. To improve overall performance, pick two elements at random, and then compare the two to make a decision.
This yields results much better than a purely random choice, and it minimizes the amount of expensive decision making necessary to get that improved result.
The important implication of this result is that even a small amount of choice can lead to drastically different results in load balancing.
Paper from Michael Mitzenmacher, 2001