Personally I think there's a fairly large cultural disconnect between shops running open-source stacks and shops running Oracle, so such a comparison would not influence user choices much. But it would be very interesting for an implementor like me: if Oracle is faster, which is likely at this point, it would set a performance point to beat.
As a final fun fact, Keyspace is currently running on top of BerkeleyDB, which is an Oracle property =) But not for much longer, I'm very unsatisfied with BDB, it doesn't live up to its image of being an industrial strength btree store (it sucks).
What do you dislike about Berkeley DB? I've been using it for a little while for a pet project, and it seems decent. I'd like it to support range-based join cursors, but that's not a major complaint. Granted, I haven't yet had to deal with database recovery, so I don't know how it stacks up there.
Also --- what do you plan to replace it with? I know that AllegroCache started out based on BDB, and ended up not using it. The Franz guys said performance wasn't up to par, and I always wished for more details.
Personally I think there's a fairly large cultural disconnect between shops running open-source stacks and shops running Oracle...
Quite possibly. Our company does use a lot of open source in the Java & Perl layers, but we're pretty tightly locked in to Oracle on the backend. I think it may be true that the sheer vastness of Oracle tempts you into believing you can solve every problem the Oracle way, which is why IOTs came to mind. I'd love to experiment with other options, myself, but its difficult once everything you do is predicated on it.
