The relational database can submit objects rapidly by means of the Redo log and also can record the operation process and the operation content. The most important method to realize database replication is transmitting and writing changed data in the disaster backup end through the analysis of the Redo log and the capturing of the changed data contents. The structure of Oracle Redo log file database is analyzed in this paper, and the header information of the log file is also stated. Considering the technology of Redo log file head positioning, the paper presents a Redo log database mining algorithm based on data block. The paper also validates the feasibility and reliability of the technology of the Redo log mining through the comparative analysis. Research direction is discussed at last.