a single binary is nice, but how to scale?

here is the overview of the Timeplus Cluster https://docs.timeplus.com/cluster#overview

basically all the cluster nodes are deployed with the same binary with extra configurations.

Interesting. How does the performance look like?

Micheal, the author of the project did some performance test and he might annouce it later, but one of his first test show Async Insert with Rate: 3422488 rows/second, but I beleive after optimization and tuning, it could be much faster.