I will add it. For example, the amount of data in my cache database has reached 10Tb, I have imported this 10Tb data into hadoop hdfs, and then analyze it through hbase. However, business data continues to be generated. The cache does not support direct docking of hdfs to store data. Therefore, I need the same cdc record table as oracle to store all dml and ddl operations. I only need to extract the data from the cdc record table. Do the corresponding operation in hdfs. The journal records the information of the global. If it is out of the cache, the global information is useless. So many times we need data at the table level.