Frequent pattern mining has a critical role in mining associations, sequential patterns, correlations, causality, episodes, multidimensional patterns, emerging patterns, and many other significant data mining tasks. With the exponential growth of available data, most of the traditional frequent pattern mining algorithms become ineffective due to either huge resource requirements or large communications overhead. Cloud computing has proved that processing very large datasets over commodity clusters can be performed by providing the right programming model. As a parallel programming model, MapReduce, one of most important techniques for cloud computing, has emerged in the mining of datasets of terabyte scale or larger on clusters of computers. Converting a serial mining algorithm into a distributed algorithm on the MapReduce framework is not necessarily difficult, but the mining performance can be unsatisfactory. In this paper, we propose a method which finds all frequent item sets by using just two MapReduce phases in a time and communication efficient manner. We demonstrate experimental results to corroborate our theoretical claims.