Recommendation on schema design

I am a newbie to Influxdb & time series db.

here is my use case to test and evaluate.

storing device metrics… have 1Million device ( device id 1,2…1M for simplicity) … Assume each device will be capturing around 50 different values ( lets just say I can keep them all as fields)

Going by the docs, if I tag device_id , then it would follow the path of High Cardinality series ( assume device id would be one of the many tags…total tags<=6)

if I split each device into its own measurement, then its considered sub-optimal schema ( ref: )

if I dont tag device_id, then I lose the capability of querying data by device…

I am confused on how this gets handled. looks very straight-fwd situation.

any insights pls?