HBMR ("HBase Map/Reduce") is a simplistic library for doing mapreduce on HBase tables.
Key principles:
- input and output are hbase tables;
- all data being processed are represented as strings;
- two kinds of mapreductions:
- a) map and then reduce; make sure you have a monoid that you reduce;
- b) just map, no reduction: meaning you most probably do not even produce anything during map phase - or maybe you do.
The purpose of this library is this:
- a bunch of people just don't understand what map/reduce is, and either are lost,
or invent their own math; that's sad;
- the existing API is approximately 42 times more complex than it should or could be.
Note. You need Java 1.6, but very soon I'm going to switch to 1.7 (with string switches).
I owe you a good wiki page that explains everything in pretty easy terms and short sentences.
11/23/2011
Vlad Patryshev