A Dynamic Fault Tolerant Threshold based Replication Mechanism in Distributed Environment

Abstract

File replication is a well known technique in distributed systems to address the key issues such as scalability, reliability and fault tolerance. For many years, file replication in distributed environment has been researched to enhance and optimize the availability and reliability of the entire system. An effort has been made in the present work to propose a file popularity based, adaptive, on demand, reliable and comprehensive file replication approach. The proposed approach introduces a File Replication Server (FRS) that is responsible for ensuring the file availability, in order to fulfill the file request. This work proposes an active replication mechanism i.e. “Threshold Based File Replication” (TBFR) model that provides file replication, access and performance transparency to the system, thereby ensuring the replication decisions about the files. TBFR replicates the file, from one node to the other node, when the total number of request for a particular file reaches the threshold value. To verify the proposed approach, TBFR is implemented on JAVA platform. TBFR model creates file replica on the basis of file popularity. The simulation results show that during high file request scenario for a particular file, TBFR replicates hot files on other FRSs dynamically and redirects the file request in a transparent manner, thus reducing the request completion time by about 28% - 48% as compared to FTP and single FRS. The Calculus of Communicating System (CCS) for the proposed model is compiled on CWB-NC tool and observational equivalence is proved for the proposed file replication and no-replication model.