Optimal data replica placement in large-scale federated architectures

Abstract

Federated architectures are largely adopted for the design and implementation of large-scale ICT infrastructure by re-using existing legacy systems and integrating novel components and systems by means of proper middleware solutions. In these architectures, data needs to be replicated and deployed in multiple parts of the infrastructure so that it can be closer to the requesting entities, and available despite of crashes and network partitions. It is challenging to perfectly determine the required number and location of the data replicas within such infrastructures. This paper provides a solution to this issue by modeling it as a multi-objective problem and resolving it as a non cooperative game. A preliminary assessment of this solution is provided by means of simulations.