AppScale is capable of running a variety of datastores when initializing. Each datastore implements the AppScale DB interface, which maps to App Engine applications. See the AppScale database comparison paper for more information on the database API.
A BigTable implementation in C++, Hypertable runs on top of Hadoop and has a primary goal of being high performing.
MongoDB is a document store best used for append-heavy operations. It is schema-free and allows for dynamic queries.
MemcacheDB uses the exact same API as MemcacheD, a distributed memory cache, and provides persistence via BerkeleyDB.
A popular shared-nothing relational database, MySQL Cluster is used as a key-value datastore in AppScale.
All these datastores are automatically deployed and configured on any infrastructure capable of running Ubuntu.