Hi! I was wondering, is there any easy-enough way to estimate Java heap usage for streaming upsert? For example, if I have 300 million primaries keys, is there a way to estimate how much heap space I’ll need?

It would depend on the size of each key. Also, there are factors like how many partitions and how many servers. I’d simply write a small java code and put the keys in a concurrent map to get some idea.

That makes sense - right now our keys are strings. If we hashed them to longs, theoretically that should take up less space right?

yes

That’s great to hear