Hacker News Reader

Concurrent garbage collectors of recent JVM versions (ZGC, Shenandoah) can give you sub-millisecond pause times, i.e. GC pauses are not really an issue any more for typical analytics query workloads. The price to pay is reduced throughput, but a scale-out architecture like Pinot makes it easy to make up for that by adding another node, if needed.

I know of at least three databases written in Java:

HSQLDB:

https://en.m.wikipedia.org/wiki/HSQLDB

H2:

https://en.m.wikipedia.org/wiki/H2_(database)

PointBase:

https://en.m.wikipedia.org/wiki/PointBase

I have briefly used either one of the first two, I think it was HSQLDB, with Java, years ago.

PointBase was created by Bruce Scott, a co-founder of Oracle, and a few others, including an architect of Sybase. This was in the early days of Java.

All three still seem to exist, according to the links above.

Neo4J

Apache Cassandra would probably be the most notable one (outside of Kafka etc).

Yes. And this is the reason ScyllaDB exists