In distributed tracing, sampling is frequently used to reduce the number of traces that are collected and stored in the backend. This is often desirable because it is easy to produce more data than can be efficiently stored and queried. Sampling allows us to store only a subset of the total traces produced.