Solution 1 :

I did similar work several month ago, hive-testbench can be an option. Check the README.md about how to make it happen.

You need to configure fs.defaultFS in $HADOOP_HOME/etc/hadoop/core-site.xml to your AWS S3 bucket, the data will be generated in AWS directly.
Pass data scale parameter to ./tpcds-setup.sh to generate date with different scale.