On the propagation of updates in distributed replicated systems

We analyse the performance and dependability of protocols which implement `sequence consistency´ in distributed replicated systems. Sites send messages to each other from time to time, passing information about the update requests that have been received. The recipient of a message is chosen according to some probability distribution which may depend on the sender. The quantities of interest are (a) the average interval between receiving an update request and being able to execute it on the original site, subject to the consistency requirement, and (b) the average time it takes to bring all sites to a consistent state after the arrival of an update