Hi, For the casual developer without academic parallel computing theory experience to draw from Hadoop's benefits are apparent but hard to implement. Design patterns?
Pig's use of tuples, bags and data atoms are more intuitive for newbies and feel like MapReduce object return types. Click-stream or web graph data can be opened and distributively filtered over namenodes by those with SQL familiarity. JSON-type formatting means you can also leverage any Javascript web presentation frameworks with the output. Brilliant Yahoo!. Peter W.
