Running a typical ROOT HEP analysis on Hadoop MapReduce

We evaluate how a typical ROOT High Energy Physics (HEP) analysis can be executed on Hadoop MapReduce. We take into account several aspects and we propose a method to perform the analysis in a completely transparent way to ROOT, the data and the user: the goal is to let ROOT run without any modifications and to store the data in its original format. The solutions which has been found to solve the encountered problems can be easily ported to any HEP code, and in general to any code working on binary data relying on independent sub-problems like HEP particle collision events. We tested the method by running an analysis code for the top quark pair production cross section measurement with the ATLAS experiment at the Large Hadron Collider in the CERN laboratory.