Building replicated database systems using distributed shared memory

Abstract

Current trends in main memory capacity and cost indicate that in a few years most performance-critical applications will have all (or most of) their data stored in the main mem- ory of the nodes of a small-size cluster. A few recent research papers have pointed this out and proposed architectures tak- ing advantage of clustered nvironments aggregating power- ful processors equipped with large main memories. This position paper proposes yet another approach, which builds on Distributed Shared Memory systems (DSMs) introduced in the early 80’s. We introduce the idea of the dsmDB, dis- cuss how its architecture could be organized, and elaborate on some of its algorithms. We conclude the paper with a discussion of some of its advantages and drawbacks.