Thanks for your comments and your suggestions.

I wasn’t intentionally swaying the votes towards SSIS, this was actually based on a real dataset that I needed to process, essentially social media data (tweets, facebook posts, etc.), so its fair to say that in this particular real-world scenario, and using the technologies out of the box (with minimal configuration), SSIS has managed to outperform Hadoop by a considerable amount.

I think in a scenario where you are trying to reduce data across multiple nodes, then Hadoop is definitely the weapon of choice, since SSIS does not have a native way of grouping results across multiple instances of the service, also, if you are trying to “scale gracefully”, then again Hadoop will win hands down.

That being said, SSIS has a much healthier start time, and seems to executes mapping operations rapidly, additionally, I believe that reducing data on one instance of SSIS will be faster than on one instance of Hadoop.

But then again, with Hadoop 2 being released, and the updated Map/Reduce engine, this experiment could be considered outdated.

Regarding your suggested scenario, its definitely a good idea to give them a shout, and if I had some free time I might give it a go!

Cheers!

then at this time you could raise some good conclusion.