Sampling metric data in Application Insights

Question

The documentation page for Azure Application Insights sampling mentions the following:

A semi-colon delimited list of types that you want to be sampled. Recognized types are: Dependency, Event, Exception, PageView, Request, Trace. The specified types are sampled; all instances of the other types will always be transmitted.

Does it mean that Metrics data (Performance counters) are always transmitted? Doesn't it make sense to sample metrics before all other types?

Answer 1

For Metrics it makes sense to aggregate and then send. This is how Performance counters work (one value per minute).

For custom metrics (TrackMetric) usually some layer aggregates (calculates average, min, max, etc.) and then call AI SDK to track aggregated value.

For these scenarios sampling is not needed.

In parallel AI team is working on layer which will take care of aggregation. So it would be possible to track individual values.