Tuesday, January 22, 2008

MapReduce: is it really "a step backwards"?

Some database people are not too happy about MapReduce starting to win the hearts and minds in academia.

The rhetoric smells like yet another religious war. This is where I spit my morning beverage on the keyboard:

Given the experimental evaluations to date, we have serious doubts about how well MapReduce applications can scale.

Furthermore it seems like they ignore that MapReduce tries to solve a very different problem than an RDBMS. One, for which a traditional DBMS doesn't provide a good solution right now.

MapReduce: A major step backwards by David J. DeWitt and Michael Stonebraker

via Artima.com

1 comment:

Michael L Perry said...

There's a good decomposition of the argument on Good Math, Bad Math.