A Dynamic Strategy for Large Data Replication in Heterogeneous Network Topologies

Abstract

In this paper, we propose an efficient method for replicating large data volumes across servers and clusters (endpoints) having geographically heterogeneous locations. The endpoints may be placed in countries with different connectivity environments while the data is as large as 5 to 10 Terabytes. Our solution relies upon a dynamic algorithm which finds the optimal paths for the data replication process from the initial node to the rest of the endpoints. In the diffusion network the edges are valued so as to maximize the utility of stable connections possessing low latency and high speed. The dynamic assessment of utilities asserts our solutions' effectiveness in topologies undergoing frequently changing connectivity conditions.