A new quorum-based scheme for managing replicated data in distributed systems

Abstract

We propose a new quorum-based scheme for managing replicated data in distributed systems. We first introduce a concept called difference pair to establish the basics for cyclic read-write coteries. A simple and efficient model is then presented to facilitate the construction of read-write coteries. The read-write coteries generated by the model are strictly symmetric. Our model can be applied to an arbitrary number of copy sites. More importantly, by introducing a parameter in the construction model, our scheme offers the flexibility of adjusting the sizes of read and write quorums. Such flexibility allows read and write quorums to be readily tailored for each individual data item according to its own request demand. Enhancement of data availability is also addressed by our model.