Tail Sampling

Tail Sampling is the process of selecting a percentage of the most extremem events or traces in an application, such as errors, timeouts or high latency requests, and discarding the vast majority of other events and traces.

Tail sampling enables you to focus on the outliers, the weird stuff that doesn't happen often, and ignore all the events and traces that are "business as usual". The main goal of tail sampling is to manage the cost of processing and storing all your telemetry data without losing visibility on the anomalies in the datasets. It enables you to zoom in on the rare events that can have a big impact on your system's performance or reliability.