Ask HN: Distributed, Persistent Hash Set?

Hey all, I have a use case that sounds relatively simple, and it's scoped very specifically. I need add/exists functionality on a large, distributed dataset (1 trillion entries at 20B/entry) at high throughput (1MM ops/sec) and low latency (single digit ms). Operational simplicity is important, so HA and easily scaling up/down should be built-in. Ideally it isn't too expensive either (100k/year ballpark would be great).

Does such a technology exist? Stuff like Cassandra or DynamoDB are heavily used, but I've had trouble operating Cassandra in the past and DynamoDB is too expensive for this use case. There are dozens of other distributed databases that have way less adoption that I'd be worried about from an operational complexity perspective. Anyone have a good experience worth sharing?

Thanks!