Abstract: Rafi consensus algorithm is a log replication consistency algorithm. Aimed at the problems of split voting and excessive load of leader nodes in Rafi algorithm, a high-performance Byzantine fault-tolerant consensus algorithm based on trust value, T_Raft consensus algorithm is proposed. Based on the idea of multi-node sharing, a parallel log replication method was proposed to increase data throughput. A node trust model was proposed to improve the voting election mechanism and log replication process, effectively detecting Byzantine nodes and enhancing the security of the algorithm. Experimental results show that, compared with the Rafi consensus algorithm, the T_Raft consensus algorithm can ensure high efficiency while having Byzantine fault tolerance, and it also greatly improves the speed of elections.