In my previous post, I introduce Map/Reduce model as a powerful model for parallelism. However, although Map/Reduce is simple, powerful and provide a good opportunity to parallelize algorithm, it is based on a rigid procedural structure that require injection of custom user code and therefore it not easy to understand the big picture from a high level. You need to drill into the implementation cod
![Parallel data processing language for Map/Reduce](https://cdn-ak-scissors.b.st-hatena.com/image/square/eaf4da596e9194d5775d56ce811646be963b5034/height=288;version=1;width=512/http%3A%2F%2F3.bp.blogspot.com%2F_j6mB7TMmJJY%2FSDmxuepaCMI%2FAAAAAAAAABc%2F6_DNic8AWmc%2Fw1200-h630-p-k-no-nu%2FP.png)