Indexing strategy

Index the paths you filter and join on; skip the rest. Each index is a persistent B-tree that speeds up reads but adds work to every write and takes disk space.

• Index high-selectivity predicates (pnr, email) — they turn scans into lookups.
• Use a unique index where a field must be one-per-document; it also enforces the constraint.
• Index nested paths and array elements you query (passenger.lastName, flights[0].from).