A Comparison of Replication Strategies for Reliable Decentralised Storage

Abstract

Distributed hash tables (DHTs) can be used as the basis of a resilient lookup service in unstable environments: local routing tables are updated to reflected changes in the network; efficient routing can be maintained in the face of participant node failures. This fault-tolerance is an important aspect of modern, decentralised data storage solutions. In architectures that employ DHTs, the choice of algorithm for data replication and maintenance can have a significant impact upon performance and reliability. This paper presents a comparative analysis of replication algorithms for architectures based upon a specific design of DHT. It presents also a novel maintenance algorithm for dynamic replica placement, and considers the reliability of the resulting designs at the system level. The performance of the algorithms is examined using simulation techniques; significant differences are identified in terms of communication costs and latency.