Write amplification measures how much physical write work a logical change triggers. Each index is a sorted copy of selected columns, so inserting a row writes the main table plus every index; updating an indexed column pays the same cost. section 7.6
Booking table versus audit log
A Booking table with eight indexes can turn one INSERT into nine disk writes. That trade-off may be sensible when bookings are read constantly and created rarely, but copying it onto an append-only audit log makes every write nine times slower for an infrequent search. section 7.6
Use these questions when reviewing a schema:
- How many indexes does each INSERT or indexed UPDATE maintain? section 7.6
- Does the read frequency justify the write penalty? section 7.6
- Are foreign keys indexed, and will deletes or newer columns create hidden scaling costs? section 7.6
@@index([startTime, endTime, status]) @@index([userId, status, startTime]) @@index([userId, endTime]) @@index([eventTypeId, status]) @@index([fromReschedule]) @@index([userId, createdAt])
Inspect the concrete index definitions and compare their covered columns with the application's common filter shapes. section 7.6

