Even if all jobs take less than 14GB data as input, but the sum total of your data is far greater, it is easier to use Hadoop as a unified single filesystem for

Even if all jobs take less than 14GB data as input, but the sum total of your data is far greater, it is easier to use Hadoop as a unified single filesystem for storing data rather than relying on multiple different machines that you manage by hand. Even if you treat Hadoop simply as a multi-machine filesystem, it automates the job of managing which files are where, that you inevitably have to manage by hand if you go with storing your data on multiple machines sans some kind of filesystem interface.