Data Replication Techniques for Data-Intensive Applications

Abstract

Abstract. Several data replication techniques have been developed to support high-performance data accesses to the remotely produced scientific data. Most of those techniques, however, do not provide the replica consistency because the data replica is just periodically updated through the remote clients. We have developed two kinds of data replication techniques, called owner-initiated replication and client-initiated replication. Our replication techniques do not need to use file system-level locking functions so that they can easily be ported to any of file systems. In this paper we describe the design and implementation of our two replication techniques and present performance results on Linux clusters.

Adaptive Replica Synchronization for Distributed File Systems

“…For the sake of showing the applicability of this mechanism, they also investigated that their system is able to achieve a smooth tradeoff between performance and availability [24]. No et al affirmed that data replication techniques can benefit to support better data accesses to the required scientific data and proposed ownerinitiated replication and client-initiated replication mechanisms to provide data consistency [25]. Moreover, recent works in a replicated parallel file system [19], [20] have demonstrated that there are a great room for improving I/O throughput and data durability and availability by employing data replication and idle storage.…”