Twitter released FlockDB as open-source software, a distributed graph database built for wide but shallow relationship networks.
FlockDB was designed to manage large numbers of simple connections, such as who follows whom or which users are associated with particular objects. Rather than optimizing for deep, multi-step graph traversals, it emphasized fast set operations across a very large network. That made it a practical fit for Twitter’s relationship-heavy data model.
FlockDB was used with Gizzard, Twitter’s distributed datastore framework, and was released under the Apache License. Twitter’s open-source infrastructure work also included projects such as Finagle and the Bower package manager. A common mix-up is Cassandra: Twitter used and contributed to distributed storage technologies, but Cassandra is a wide-column database, not the graph database named in this question. FlockDB is no longer actively supported by Twitter.