Good question. I would like to know about this as well :-) I run a mini database with datahike file backend. I have the same logic also with mongo db and on mongodb everything is much faster. This is weird since with mongodb there is an additional encoding layer involved. My worst experience with file store so far is fragmentation that occurs over time db size grows x 10000 due to updates in data, even with history disabled. My writes are slow ( I have lots of small transactions that I could rewrite to be way fewer). So the question of other backends is a very important decision.

Hey, yeah we're currently trying to get the synchronous interface back into Datahike and after that we would like to add new backends as well since some of the protocols needs to be adjusted. We had the idea to include RocksDB and H2. Would that be of interest?

Regarding the performance we're moving away from the hitchhiker-tree to the persistent set since the performance problems with upserts are limiting the transaction times.