I am using Timescale DB to store measurements from many devices. Since most devices so far had a similar set of metrics, I used a hypertable with many columns (one for each metric).
However, some new devices will have to support different metrics, and there are also derived values which I need to calculate and store for each device.
So it seems I have to store these values as separate tables, or use the normalized approach (timestamp, device, metric_type, value) if I want to have a general way for saving time-series data?
Is the "many columns" schema approach wrong for these kinds of measurements, or am I doing it wrong?