berkeleydb

The storage backends are pluggable, but most commonly used ones are BerkeleyDB and read-only (for data built off-line).

If rumors are to be believed, SimpleDB is really just Erlang + BerkeleyDB with some cpu limits (which makes sense since BerkeleyDB really only gives you access methods and raw storage).

But the short version is nginx reverse proxy into hunchentoot, with elephant (and a BerkeleyDB datastore).

Maybe it's because relational databases got the nice consistency features first; Interbase certainly predates BerkeleyDB.

BerkeleyDB is awesome, but its license may not be suitable for many organizations.

Maybe I was thinking of InnoDB or BerkeleyDB or somesuch and just conflated them all together...

The data they store is so simple and non-relational that they should be making heavy use of memcached and/or BerkeleyDB and/or flat files and/or something else extremely fast and simple.

As for the former, key-value stores from the basic BerkeleyDB (though I think there's more modern and better software now) up to three-dimensional bigtables are really conceptually simple when it comes to your data.

Sleepycat did it with BerkeleyDB (they were bought by Oracle, though, and I don't know if Oracle changed this arrangement), the same goes for MySQL (which, of course, is now owned by Oracle, too).