Link Details

This is a badly written and researched article by one of the GridGain engineers, inspite of the author's claim of neutrality.

Former GridGain engineer. And let's agree that it's neutral

And as always voted up to the main page only by GridGain developers.

I think this short overview is well balanced. If I were to write it - it would be a lot more scathing... Over hundreds of active users I have never seen Hadoop compared or competing with GridGain. We've seen plenty of GigaSpaces, some Terracotta - but never Hadoop. I have pretty good theory on why - but that's another topic.

Nikita Ivanov.
GridGain - Grid Computing Made Simple.

> And let's agree that it's neutral [ ... ]

Why? Because you, the author, say so?

> I have never seen Hadoop compared or competing with GridGain.

That's perhaps because what you call MapReduce is not what the rest of the world calls MapReduce, but something different, and thus impossible to compare.

http://en.wikipedia.org/wiki/MapReduce

Hey calm down. I don't like to see you both fighting.... Hmm but may be it's a good Idea to yourtube this fight :)) I think I would get a lot of comments and views on that ;). It does not matter how close is you to the certain rule or schema. It's all about usability and simplicity. I personally don't care if someone hides real map/reduce and gives me the way to execute on grid my code without writing map/reduce code. I will call it map/reduce if vendor says so.

The incorrect parts of the article are:
1. Splitting the input data and splitting computation work are isomorphic to each other. It is trivial in Hadoop to have splits that don't correspond to literal inputs.
2. GridGain doesn't implement Map/Reduce. It is a traditional distributed work queue manager that happens to have a method named "map" and a method named "reduce". It lacks any distributed sort or any capability for data intensive applications, which are the two defining characteristics of Map/Reduce. I wish the GridGain team would stop trying to confuse their potential customers by referring to their system as Map/Reduce.
3. Hadoop runs on Linux, Windows, Mac OS, and Solaris.