Where are readings stored and how long do I keep them? — answered by Engineer Saeed Baatiah
Readings accumulate faster than anyone expects: one device sampling every minute produces over half a million readings a year. The practical answer is summarisation rather than deletion: keep full detail for a short window, then hourly averages, then daily averages for the archive. That keeps you able to compare this month against last year without a storage bill that balloons.
Work out the volume before it surprises you: one device sampling every minute produces 525,600 readings a year. Ten devices means over five million rows, fifty means twenty-six million. That is not technically impossible, but it changes the database design, the storage cost and the dashboard’s speed — which is why it is calculated at the start rather than when the screen slows down a year later.
The answer is tiered summarisation rather than deletion: keep the full raw reading for a short window long enough to diagnose any recent fault, then hourly averages for several months, then daily averages for the long archive. That keeps you able to compare this month’s consumption against the same month last year — the comparison decisions are actually built on — without a storage bill that balloons.
- One device every minute = 525,600 readings a year
- Tier it: short raw window → hourly → daily for the archive
- Keep enough to compare this month against last year