On the power of reconfiguration in fault-tolerant distributed databases

Dynamic reconfiguration is used to significantly increase fault-tolerance in replicated databases. This is illustrated for both unstructured and logically structured replicated databases. Previous work on reconfiguration was mainly concerned with changes within a particular replica management protocol. The authors extend this approach to apply reconfiguration from one protocol to another. A generalized reconfiguration mechanism is proposed for designing large and scalable distributed databases. Reconfiguration not only ensures a high degree of fault-tolerance but also provides high performance. Finally, the generalized reconfiguration mechanism can be used by the database designers to dynamically adapt database performance and availability to changing application needs and underlying network conditions.<>