Performance considerations for metrics

Also see Performance considerations for logging.

Process samples before converting them to Metrics

emit’s infrastructure for metrics is largely concerned with emitting them, not with processing them. If you need to perform normalization or aggregation, do it before your samples reach emit. This way you can define a stronger data model around the kinds of metrics your application produces, instead of trying to work with emit’s loosely typed bags of properties.