As Facebook started to build their new messaging system last year, they had to decide what infrastructure to use. “[The Facebook infrastructure] wasn’t really ready to handle a bunch of different forms of messaging and have it happen in real time,” said Facebook engineer Joel Seligstein as interviewed by The Register.
After testing different platforms, it turns out that open source distributed database software HBase was the final decision. Facebook used to use MySQL for message storage and Cassandra for inbox search. Now Facebook is using Hbase for e-mail, chat, messages, and SMS.
HBase was built by Powerset, a semantic search engine company acquired by Microsoft. HBase is used by Facebook to store text and metadata for messages and the indices needed for them. Before Facebook started to roll out the messages, they have been juggling 15 billion on-site messages per month and 120 billion chat messages.