0 / 60 seg.

Rather than adding a Secondary Sort to a slow Reduce job, it is Hadoop best practice to perform which optimization?