It would be nice to be able to do this based on primary key/underlying partition. But I can see how technically it’s probably easiest to implement at the table

It would be nice to be able to do this based on primary key/underlying partition. But I can see how technically it’s probably easiest to implement at the table level. Writes are priced higher, which means there’s probably a crossover point where the lower storage offsets the write cost (especially if you have to essentially write twice, once to a current table and again to this table). Still the economics feel like they make sense in spite of the technical annoyance of needing to search two tables, likely split by a point in time, and needing to move data between tables, or write to both. Would it be better to migrate data between tables at some point? Or write to both at the same time, TTL expire your current table, and then in your app know when you need to search the archive table (say for something older than 6m).