NCRG: A Node Classification Method Based on Updated Risk Graph under Hybrid Replica Consistency Strategy

Hybrid replica consistency strategy can effectively deal with the replica consistency problem in dynamic cloud storage. These strategies usually classify nodes according to a feature that affects replica consistency, and then update replica according to the node classification. However, since the replica consistency in dynamic cloud storage is restricted by many factors, the classification method that consider a single factor may affect cloud storage availability. To deal with this problem, this paper proposed a node classification method based on replica update risk map (NCRG for short), which can be used to design a new hybrid consistency method, aiming to improve the replica update timeliness and update probability in dynamic cloud storage. NCRG considers many factors that affect replica consistency, and divides these factors into two types: node service capability and node service willingness. Then, this method builds a replica updating risk map according to this attribute classification. Finally, the nodes are divided into two types: stable nodes and variable nodes according to the position of nodes in the risk map. Extensive experiments show that NCRG can obtain superior performances in replica's probability of receiving updates successfully around 5% on average than other similar methods.