Wednesday, February 13, 2008

Database verses MapReduce

You might have noticed that there's been a discussion by some dbase guys about how MapReduce is a step backwards.

The article is a little shortsighted IMHO. It's quite well known that databases DO NOT SCALE. Trying running a distributed database on 10,000 machines. In order to get the scalability of facebook, google etc you have to take the data OUT of the database, and generally dump them in memcached.

In any case, here's a nice response to the above article by someone who works for (but not speaking for) google.

No comments: