Universal Replication Architecture

The goal of this proposed research is to develop a unified data replication architecture and toolkit for large-scale distributed services. This architecture and toolkit would represent a single set of mechanisms on which a broad range of data replication systems can be built for wide area, enterprise, grid, web, and mobile data systems. The core hypothesis is that many existing systems are ``special cases'' of a more fundamental underlying protocol, but that they are superficially inconsistent because they embed specific policy assumptions in their implementation mechanisms. If this hypothesis is true and the proposed work successful, this research will result in a set of unified, flexible, and orthogonal mechanisms for consistency, for placement, and for topology that (1) significantly reduce the costs and improve the performance and availability of replicated data systems and (2) reduce the development effort and time-to-market for constructing new systems for new environments.

The talk Towards a Unified Theory of Replication outlines this vision and some early steps towards realizing it.

Key ideas for achieving this goal include




Mike Dahlin's other research projects