DRF-FTS: A Dynamic Replication Factor Replication Scheme Based on Fault-Tolerant Set

Cloud storage systems use data replicas for fault tolerance, however, this approach does not provide enough protection for data when a correlated failure occurs within the cluster (e.g., a power outage that causes some machines to stop working) and it is almost let to lose data. The traditional random replication strategy cannot provide high data reliability, and the traditional FTS approach does not consider the impact of data heat on the whole system. In this paper we propose a dynamic replication factor replication scheme based on fault-tolerant set named DRF-FTS. In our approach we classify data into different types based on data popularity and use fault-tolerant sets generated by the algorithm to restrict the placement for different types of data. Experimental results show that DRF-FTS has high data reliability (much higher than the widely used random replication strategy) and improves read performance by 14% compared with the traditional FTS approach.