For a couple similar use-cases (write rarely, read many times, JSON data) I ended up using actual JSON files, on disk. Time to read and parse them is very small

For a couple similar use-cases (write rarely, read many times, JSON data) I ended up using actual JSON files, on disk. Time to read and parse them is very small compared to asking Redis or PG for it, and this benefits from the Linux disk cache, which makes reading from those most-unchanged files almost the same as reading it from memory.