Abstract
In dynamic cloud, more replicas can reduce access delay and balance the node load, but more difficulty of replica consistency and vice versa. To deal with this problem, this paper designed and implemented a dynamic replica placement method based on the node's credit risk, named RPCC, aiming to use replica placement to balance the relationship between system performance and replica consistency. To achieve its goal, RPCC uses the decentralized adaptive manner to dynamically create and place replicas. This method selects files with high popularity and low update rate to create replicas when creating new replicas. Meanwhile, it plots the credit risk map according to node load, node degree and access delay. RPCC selects optimal nodes to place replicas according to the credit map. Extensive experiments show that RPCC can obtain superior performances in consistency update probability increase around 5% on average, in access latency decrease around 10% and higher load balance than other similar methods.