Failure recovery mechanism in neighbor replica distribution architecture

Abstract

Replication provide an effective way to enhance performance, high availability and fault tolerance in distributed systems. There are numbers of fault tolerant and failure recovery techniques based on replication. These recovery techniques such as Netarkivet's data grid and fast disaster recovery mechanism for volume replication systems were implemented in two-replica distribution technique(TRDT) or primary-backup architecture. However, these techniques have its weaknesses as they inherit irrecoverable scenarios from TRDT such as double faults, both copies of a file are damaged or lost, missing of the content index in index server table and index server has generated checksum error in content index. In this paper we propose the failure recovery based on the Neighbor Replication Distribution technique (NRDT) to recover the irrecoverable scenarios and to improve the recovery performance. This technique considered neighbors have the replicated data, and thus, maximize the fault tolerant as well as reliability in failure recovery. Also, the technique outperform the TRDT in failure recovery by reducing the irrecoverable cases in TRDT. It also tolerates failures such as server failures, site failure or even network partitioning due to it has more the one back up or replica. © 2010 Springer-Verlag Berlin Heidelberg.