Replex

Replex is a novel concept that can be applied to any distributed database. We introduce the data replication unit that we call a replex, which combines replication and indexing. We then fundamentally change the replication paradigm by replicating data to replexes, as opposed to exact data replicas. In doing so, distributed databases can greatly reduce the overheads of supporting secondary indexes, which are crucial to supporting rich functionality.

An open source implementation of Replex can be found integrated into the Corfu database.

Publications

  • Replex: A Scalable, Highly Available Multi-Index Data Store
    Amy Tai, Michael Wei, Michael J. Freedman, Ittai Abraham, and Dahlia Malkhi.
    USENIX Annual Technical Conference 2016
    (ATC ’16) Denver, Colorado, June 2016. [paper as pdf] [presentation]
    Awarded Best Paper