A New Data Placement Approach for Heterogeneous Ceph Storage System

In the condition of heterogeneous Ceph storage cluster, the data distribution is imbalanced due to the pseudo-randomness of the CRUSH algorithm. In addition, the CRUSH algorithm only considers the nodes storage capacity to determine data storage location without considering the different ability of nodes in data processing, which will reduce the performance of cluster. A new data placement approach for heterogeneous Ceph storage system is proposed to solve these two problems. This proposed approach first adopts a multiple attribute decision-making model integrating the factors of storage capacity, CPU performance, memory size of each node, and then the probability weight of each heterogeneous node is determined by solving the proposed model to balance the data distribution. The result of series real-scene experiments shows that the proposed approach can not only improve the reading and writing performance and the fault tolerance but also make the data distribution more balanced.