We can achieve a 6x performance speedup and we can drive that to 16x by using a simpler MapReduce model to avoid data motion more efficiently than Hadoop.

FORBES: Big Data Analyzed Fast In Memory -- For Finance, Online Dating, Home Shopping Network