The document describes the transition of Formspring's social graph infrastructure from a MySQL and Memcache system to a Cassandra database. It discusses the scaling issues they faced with high growth, including slow queries and out of memory errors. Formspring built a social graph using Python, Cassandra, Pycassa and Thrift to model following, followers, blocks and blockers. They iteratively loaded user data into Cassandra and migrated reads and writes from MySQL to the new Cassandra cluster. This allowed their "Ask Followers" feature to scale to support 14k queries per minute for following counts and 40k queries for follower counts during peak times.