Flexible data replication using constraints

Motivation:

Data replication is crucial to mobile devices to extend their storage resources and achieve fault tolerance. Recent research work in the context of systems like Cymbiosis has explored how to allow users to express content-based filters to control on which devices data is made available and regularly updated. In this project we explore how to provide support for a larger number of replication policies that go beyond filtering data based on their content. The system enables users to specify filters based on the geographical location of their data as well as the latency at which they must be accessible from specific devices.

Goals:

Provide a flexible and extensible set of data filtering techniques that can meet the requirements of today's distributed storage systems and primarily their heterogeneity and dynamism. Implement assignment of data to backup nodes, dynamic detection of addition/removal of backups, load balancing among replicas, garbage collection. Use constraint logic programming to implement the identified techniques and test their performance and correctness on the Rhizoma overlay, which federates both personal computing devices (PC, laptops, phones) and private/public cloud computing infrastructures.

Contact:

Gustavo Alonso