Downsampling

[cyriltovena]

Suppose we store natively every 15s.
We can run a process to aggregate every 5 min, say, and store that too.
This speeds up queries because we add up 20x less data.
Maybe do every 1 hour, every 4 hour ?
This would be deduplicated at the point of aggregation.

[cyriltovena]

cc @bboreham