Lines of Activities
Lines of Activities
Distributed Applications
Design and development of robust distributed applications. Use of replication technologies for improving the availability and reliability/dependability of distributed applications. In an academic context, we have focused research in this topic on database replication, since it is one of the areas that poses major difficulties in terms of applying replication technologies(due to the size of state that needs to be replicated and to its specific concurrency control). The solutions that can be found in that area are therefore easily migrated to other kinds of applications. That line of research provides a general framework which has been refined in the lines of research and development described in the sequel.
Replication Protocols
Design and development of new replication protocols. The replication strategies have to be implemented using certain protocols. There are several different established types of replication protocols, but new, more specific variants can be developed. For example, to improve scalability. In the case of replicated databases: support of each isolation level (theoretical analysis), protocols that support more than one level, ...
Consistency Models
Study of consistency models. Each replication protocol will manage a certain level of consistency. Although sequential consistency is usually guaranteed, not all replication protocols maintain precisely that level. Subtle differences can significantly improve performance or scalability, while still providing an easily implementable model.
Adaptability in Replication Systems
Improvement of the adaptivity in replicated systems. Design and development of meta-protocols that permit the simultaneous use of various replication protocols (for the same consistency model). As the performance of replication protocols varies between different classes (eg, some classes use optimistic concurrency control, being much more efficient, but aborting some operations, while others employ pessimistic control, ensuring the success of any operation, but lower efficiency), such a metaprotocol would allow each application / transaction / operation to choose the management protocol, thereby improving overall performance. It will also permit dynamic swapping of protocols (that is, during operation and without causing blocking phases).
Integrity in Replicated Systems
Study of mechanisms to manage the integrity of relational databases, analyzing if they are capable to support extant inconsistencies. Establishment of recommendations for use of such mechanisms in replicated databases. Study and improvement of integrity in support of replication protocols. Use of integrity management mechanisms in other types of distributed applications.
Strategies of Recovery
Design and development of recovery protocols which can be easily integrated in the replication protocols. These protocols turn out to be necessary to re-incorporate replicas that have previously failed, or to increment the number of replicas, transferring the state with which they must initiate the management of new queries. During recovery, the normal processing of queries in the replicas must not be interrupted. The recovery turns out to be difficult when the size of state to be transferred is large (databases, for example).
Distributed Algorithms
Design and formal correctness proofs of new solutions for distributed systems.
Group Communication Systems
Development of group membership monitors and goup communication protocols. Execution models based on virtual synchrony. Prioritized communications. Support for group communication in MANETs.
Replication Middleware
Design and development of replication middleware systems, where each of the components that have been generated in the lines of r & d described above can be integrated.
Dynamic Systems
Distributed solutions for dynamic systems. Adaptation of traditional distributed algorithms in environments that are highly volatile (with regard to the workload as well as with regard to the number of nodes
RSS
