Abstract: The data distribution algorithm in Ceph distributed system only takes the capacity as the standard for selecting storage nodes, and does not consider the network state and node load of storage nodes. In the replica mode, when a storage node in the three replicas needs to be repaired, too high node load or network load will lead to a large node repair delay. To solve this problem, a fault-tolerant node selection algorithm based on Ceph (FTNSC) is proposed. Firstly, the software defined network technology is used to obtain the real-time network state and node load information as the data support of the node selection method; Then, a multi-attribute decision-making mathematical model considering the node load information is established to determine the location of the primary storage node; Finally, the artificial bee colony algorithm is used to obtain the optimal secondary storage node according to the network state and node performance with the primary storage node. Experimental results show that compared with the existing methods, the proposed method can improve the performance of data storage nodes and reduce the data repair delay by 2%-29.7%.