We were able to get INTEL Hadoop running on our servers (Cent OS 6.2) and then conduct a benchmark with INTEL's Hadoop compared to a Generic Hadoop distribution.
We see increase of 30% (which go upto 55%) on join query as the dataset was grown. We plan to provide this as an AMI for free to community to let other also validate the results. Please find attached white paper listing steps, results and learnings.
Regards
Harpreet Singh