File heat-based Self-adaptive Replica Consistency Strategy for Cloud Storage

Abstract

In cloud storage systems, replica is a key technology, which reduces access time lag, network bandwidth consumption and system unreliability. However, an inadequate replica consistency management mechanism would cause problems for cloud storage systems in the time lag of file access and network bandwidth consumption. Therefore, this paper proposes a self-adaptive replica consistency strategy which allows the system to automatically select the adequate consistency strategy according to file heat. In the calculating of file heat, this paper takes into account the effect caused by the pattern of file access time sequence in a time period and introduces a new algorithm MRFU for the calculating of file heat, by which the file used most recently and frequently in a time period is assigned the largest heat value. Experiment results show that the file heat calculated by MRFU algorithm can precisely predict file access behavior. On the basis of keeping high file access efficiency, the proposed self-adaptive replica consistency strategy effectively reduces the network bandwidth consumption. Furthermore, it is found that the proposed self-adaptive replica consistency strategy has better performance in file access efficiency and cost of network than other similar solutions by contrast tests.