Question: how well does GreptimeDB handle a large number of relatively small databases on a single instance? Assuming low concurrency in terms of access to the database, so cache thrashing isn’t a concern.

We haven’t tested that. But from another angle, we have tested concurrent reads and writes on million‑table scale, and the results were not bad.

The "database" concept is more like a lightweighed namespace. In most cases, the table count matters more.

Thanks for the quick response. Is there a known best practice upper bound for per-instance table count? Eg something you’ve tested against

I think it depends on the instance spec such as CPUs and memory. And maybe you can try 10,000~50,000 tables per instance.