The SIP server can be comprised of an engine tier and a state tier distributed on a cluster network. Engine nodes in the engine tier can process SIP messages and can read/write state information from/to the state tier. State tier can maintain state information in a set of partitions of one or more replicas which contain duplicate information. The engine nodes can be adapted to detect and report replica failures and the replicas can in turn be adapted to detect and report engine node failures. Replicas can detect faults with an engine node if the engine node fails to poll the replica for a specified period of time and can then report the failure. The engine node can detect failures of a replica when reading or writing state information and can report the failure to another replica, which can be responsible for updating the partition view to exclude dead replicas.