Dynamic data replication based on access cost in distributed systems

Data replication is a common method to reduce the bandwidth consumption and the access latency in distributed systems. The existing replication methods mostly deal with the read-only data, thus the replica management mechanism is relatively simple. Most replication methods replicate files in all nodes or two endpoints on a query path. Naturally, the larger number of replicas, the better performance it can obtain. However, these methods lead to high overhead for unnecessary file replications and consistency maintenance. This paper presents a dynamic minimum access cost based replication strategy (MAC replication strategy). MAC replication strategy takes into account the access frequency, the status of the network connection and average response time to perform optimal replication. It first selects the popular files. To each popular file, the average response time is calculated to determine which logical resource should be replicated. Then MAC replication strategy calculates an appropriate site to replicate for better shortening the response time of the data resource. Simulation results demonstrate the efficiency and effectiveness of MAC replication strategy in comparison with other approaches.