This task regards the design of a fault-tolerant data management service targeted for large and geographically distributed applications by applying replication as necessary to ensure reliability, avalability and performance. These are, however, conflicting goals and as such an essential part of the study shall consist on the identification of crucial trade-offs and on the provision of configuration flexibility when possible.

This task will be dedicated to the implementation of the data management system. As it can be forseen at this time, the system wil be composed of four modules whose dedicated functionality is meant to provide a clear separation of concerns. We start by describing the general responsabilities of each module. Then, we develop their possible refinements and interconnection in order to show how the system can accomodate the different replication strategies and solutions resulting from the research of the previous task.

The purpose of this task is twofold: to provide quantitative data to support design decisions in Model and algorithms and to evaluate the implementation resulting from Implementation.