The problem: A large scale manufacturing organization aggregates data from different sources, maintains it in a single Oracle table, and the number of records is in the order of a little over a billion. A monthly process has to fetch the data from Oracle to HDFS. The constraint: Ideally, only the difference for each month could be fetched. But, there is little to no control over the Oracle data source and there is no reliable way to identify the delta. Hence, all the data have to be fetched all the time. To give a perspective, if the table is exported as a CSV from a SQL Client (say, SQL Developer), it takes more than 20 hours to download the table. The tool: Sqoop is the standard tool used to import data from the relational database to HDFS. The solution: $ sqoop import -D **oracle.row.fetch.size=50000 --fetch-size 15000 --num-mappers 40** --table ` <schema>.<table_name> ` -connect ` <jdbc_connection_url> ` --username ` <user> ` -P --target-dir...