Consistent Hash Based Flexible Redundant Data Partition Strategy

Abstract

Because of the development of computer networks and the scattered locations of system, distributed data-storage system is widely deployed. Data partition and redundant strategy determine the space utilization, access performance and data reliability of the storage system. Based on the analysis of current data partition and redundant mechanisms, we propose a novel Consistent Hash Based Flexible Redundant Data Partition Strategy (CHAFRAS) for key-value data models. It partitions the hash ring dynamically according to the number of virtual nodes and manages data copies on the level of partitioned unit. Furthermore, CHAFRAS shortens the path of data access and simplifies the management of data copies on the basis of keeping the features of scalability and load balancing. Extensive experimental results offer an insight into the proposed techniques and suggest that they are efficient in terms of convergence time and access delay by comparing with the state-of-the-art storage system Dynamo.