Hadoop brings the ability to cheaply process large amounts of data, regardless of its structure. By large, we mean from 10-100 gigabytes and above. How is this different from what went before?