With all due disrespect, you're wrong. Go ahead and implement your solution, and you will find it falls apart. So tired of people pretending to know better, wit

With all due disrespect, you're wrong. Go ahead and implement your solution, and you will find it falls apart. So tired of people pretending to know better, without any data or real details to back it up. A "proper database" would not scale, regardless of whether it is Cassandra or Postgres. You're completely ignoring, or completely oblivious of the fact, that the entire 1000x1000 grid must be provided to every connected client. You're not going to read out one million aggregated rows by most recent timestamp per cell, from a billion rows of history, in a scalable amount of time. Please post your GitHub link that proves your solution as superior, or even viable. Make sure it includes database triggers, for which you don't explain how they would help scale the app whatsoever. Are you going to have a denormalized table containing each of the one million cells' most recent rows? All you are doing is eliminating a GROUP BY on the indexed cell+timestamp columns. It's still a million rows returned per query. Please explain how that scales. Eagerly awaiting your proven solution that defies common sense scaling logic.