Metrics
In my opinion, metrics are the most useful form of logging for day-to-day operations. Metrics are useful because we have simple numeric data. We can plot this onto a time-series dashboard and quite quickly set up alerting from the output because the data is incredibly cheap to process and collect.
No matter what you are storing, the superior efficiency of metrics is that you are storing numeric data in a time-series database using a unique key as an identifier. Numeric data allows the computation and comparison of the data to be incredibly efficient. It also allows the data store to reduce the resolution of the data as time progresses, enabling you to have granular data when you need it most at the right time and retain historical reference data without requiring petabytes of data storage.
Types of data best represented by metrics
This is quite simple: it is the data that is meaningful when expressed by simple numbers, such as request timings and counts. How granular you want to be...
