A Group-Based Replica Consistency Schema for Cloud Storage

Abstract

Replica is the most meaning of data availability, scalability and reliability in the cloud storage. There are rooms for reducing the cost of consistency by designing an effective replica management schema. To address the conflict between potential inconsistency and efficiency problem, a new group based replica consistency schema, namely GBRC, was proposed in this paper. The schema will divide replicas located on different network topology into independent replica groups and provide two kinds of consistency management mechanism for replica group and replica members of group. In the GBRC, replica group is an autonomous unit on group management level. It is an effective way to organize adjacent replica node into the same group to reduce cost of communication and complex of management. In this paper, a novel replica group construction process adapted to select adjacent nodes which meet user storage requirement to be self-consistent group. Then, all groups of the same replica consist of entire replica management system on the top level, which is effective way simplify replicas organization, management and consistency. Experiments show that replica group autonomy consistency respectively can improve consistency speed and simplify metadata management especially when replicas are geographic distributed. In addition, design a new dynamic replica consistency strategy based multi replicas consistency quality of service. The strategy dynamically adapts replica consistency number at runtime to achieve a dynamic service balance between consistency and quality of service. Evaluation results show that consistency strategy improves cloud storage responding speed while guaranteeing the user data storage’s consistency requirement.