Reconfigurable algorithms in view synchrony

A common approach to designing fault-tolerant distributed systems is to use a view-oriented process group communication service as a building block. An application that uses such a service often operates by a simple algorithm (such as a replicated state machine) for the duration of a view, but whenever the view changes, a complicated calculation is used to reconfigure the system. This reconfiguration typically involves exchange of state and merging of information from previously partitioned views. In this paper, we show how to automate the design of the reconfiguration mechanism in certain simple cases, where the reconfiguration treats all previous views symmetrically