Improving Efficiency and Performance of Distributed File-Systems

Abstract

This thesis presents a distributed file-system for the present day medium-sized network, such as those found in academia or the corporate environment. Existing servers and workstations pool their unused storage resources to form a communal share. Era sure codes provide fault tolerance and eliminate the need for replication. Middleware libraries are used to form an overlay network and facilitate object routing and location. The system presented shows improvements that can be made to present day distributed file-systems that would benefit scalability, fault-tolerance, and throughput at a marginal cost to the computers in the existing environment.