Replicated K-resilient objects in Arjuna

The design of an object replication scheme for the Arjuna distributed system is described. The design supports K-resiliency, where, in the absence of network partitions, K out of a total of K+1 replica failures can be tolerated before an object becomes unavailable. The scheme chosen uses active replication, in which each and every functioning replica of an object carries out processing. Computations are structured as atomic actions (atomic transactions). The details of how object groups are created and terminated, how a group can be evoked, and how object replicas are inserted and removed in a consistent manner in the presence of node failures are presented.<
>