Participants of our Distributed Data Management lecture already know Bigtable – Google's fast and distributed NoSQL database that has been reimplemented in the open source Apache HBase project. Being NoSQL data stores, Bigtable and HBase miss some features commonly found in relational database systems. For HBase these are e.g.: typed columns, secondary indexes, triggers, and advanced query languages [via]. More generally speaking: NoSQL databases do not support (all of) the ACID properties (students should remember them from our RDB1 lecture).

Not supporting the ACID properties means not having crucial properties needed for some high-security applications – like Google AdWords, which is one of Google's most important applications. Being used by millions of paying customers, AdWords has high demands in terms of data consistency and integrity – demands the Bigtable system cannot entirely and appropriately address. Since state-of-the-art databases like MySQL cannot deal with the large amount of AdWord's data any more, Google developed F1 which provides (1) scalability, (2) availability, (3) consistency, as well as (4) usability – and thus proves several researchers wrong that stated these design goals were mutually exclusive.

Read more about this fascinating new distributed relational database in their VLDB paper or in this (german) Heise article.