That's a strategy that can be applied to various problems. When you can use it, it's a very effective way to make use of multi-processors/machine clusters.

It seems to be particulary well suited for problems where the processing can be split into more or less independent sections. Example: most fractals can be rendered partially - each process takes on a "tile" and calculate only that part.

Recently, I've been thinking about audio processing - where parallel audio streams can be split out:

CPU 1       CPU 2          CPU 3

O source one               O source two
|                          |
+-----------+              |
|           |              |
O effect 1  O effect 2     O effect 3
|           |              |
|           +--------------+
O effect 4  O effect 5
|           |
|           |
O output 1  O output 2
[download]

Signals go in from the top and exit at the bottom (usually in small batches). Each column can in principle be taken on by a different process(or).