The rule of thumb for Druid is around 10k records / second / core. Of course, YMMV and Druid is linearly scalable and very configurable. Please keep in mind that it is optimized (and incredible good at) for write once / read many OLAP style workloads. If your data changes often, or you have a write-heavy workload that doesn’t require support for complex, ad-hod style queries, you may want to look at the plethora of other databases that are designed for that type of workload.
Please let us know how we can assist you on your Druid journey!