Epidemic Quorums for Managing Replicated Data

In the epidemic model an update is initiated on a single site, and ispropagated to other sites in a lazy manner. When combined with version vectorsand event logs, this propagation mechanism delivers updates in causal orderdespite communication failures. We integrate quorums into the epidemic modelto processes transactions on replicated data. Causal order helps establish thetransactions on replicated data. Causal order helps establish the globalserialization order on transactions. Our approach enforces serializability byaborting transactions that may cause inconsistency. In the absence of conflicta transaction can commit as soon as it is known to a quorum of sites. In thepresence of conflict, sites vote and a transaction can commit as soon as aquorum of sites vote for it. We present a detailed simulation study of adistributed replicated database and demonstrate the performance improvements.