> you should be able to put a journal mode SQLite file on a cephFS volume without any changes to SQLite. If everything works to spec, then at minimum every SQLite transaction requires two network round-trips. (Commit to journal, move data to DB.) With the HTTP-based API typical of the distributed DB layers over SQLite I linked to, that'll be one round-trip per transaction. Since network I/O time will typically be an order of magnitude or so slower than the storage I/O in the sort of system where Ceph will be in use, that means using SQLite-on-Ceph will be at least twice as slow as the alternative. That's all the single-node perspective, but I don't see that it breaks down when you start talking about multi-node consensus. Both schemes have to arrange a way to do that, so as long as you're comparing apples to apples, *that* I/O overhead should amount to approximately equal, so you can leave it factored out of the discussion. But hey, try it. It's not difficult to work up some simple benchmarks here.