Thomson Reuters interview question

How to partition a huge volume of data for distributed processing?