Database replication: if you must be lazy, be consistent

Due to the severe performance penalties associated with isochronous replication, there is significant interest in asynchronous replica management protocols. Lazy protocols currently in use either do not guarantee consistency and serializability, as needed by transactional semantics, or they impose restrictions on the placement of data and on which data objects can be updated. In this paper, we consider an alternative update protocol, based on epidemic communication, that guarantees consistency and serializability in spite of a write-anywhere capability