Sequoia RAIDb-3: a new model for data distribution and replication using commodity systems

Unprecedented growth in the amount of data generated and used in the modern world has made distributed databases an important target for contemporary research. Today, advances in distributed databases embrace a wide range of concepts and ideas including, but not limited to, fragmentation, replication, distributed transactions and distributed concurrency control. One novel idea in this area was the introduction of the Redundant Array of Inexpensive Databases (RAIDb) initially proposed by the authors of Sequoia, a Java- based clustering middle-ware framework. On a conceptual level, RAIDb is similar to RAID arrays of disks; however, in contrast to traditional RAID, RAIDb utilizes an array of individual databases. The objective of RAIDb is to provide improved performance and fault tolerance relative to a single database while preserving the abstraction of a standard SQL DBMS. This thesis extends the functionality of Sequoia and proposes a distribution model based upon full horizontal fragmentation. We refer to this new design as RAIDb-3. We discuss details of the implementation and support its validity with an extensive suite of test cases