Hi JB,
I believe the difference in ease of use is essentially due to what initial focus was given to the framework.
I created JPPF specifically because toolkits like Globus or Hadoop were such a pain to get started with, and at the time, I was looking for a compute grid framework that would allow me to run my experiments with frequent changes in the code or in the simulation parameters. If it takes longer to deploy my changes than to write them, then it's a no-go for me. So JPPF was made with ease of use in mind from the very start.
On the other hand, JPPF and Hadoop handle different classes of problems. JPPF focuses on compute-intensive tasks (more formally it addresses the class of embarrassingly parallel problems), whereas Hadoop is made to handle and process extremely large datasets (think "Google"), hence the usefulness of a distributed file system. So here again, different focus.
Without being an expert in rendering, I do have the impression that JPPF is a better fit for this type of work, and I'm hoping this'll hold true for your project.
-Laurent