java ,sqlldr and shell script use for 10 million data

I want to upload 10 million data to oracle DB from a file which is in FTP.I am using a Java +shell script+SQLLDR combination in following way--

1. upload the 10 million data file in FTP manually.
2. run a java code
3. call a .sh from within the java code
4. the .sh removes the duplicates from the 10 million data file and creates a new file free from duplicates
5. the .sh calls SQLLDR to load the new duplic开发者_高级运维ate free file data in oracle DB

Please advice me if I am following the optimum way .thanks to all :)

I presume that you sort your data file to remove duplicates. In that case you can get better performance using direct path loading with SORTED INDEXES parameter.