In this paper, an extremely efficient architecture for the Simultaneous Localization and Mapping (SLAM) problem is proposed. This architecture depends on distributing of heavy computational tasks and large data sets among remote servers and frees the robots from any computational loads. Thus, the most widely used FastSLAM2.0 approach is parallelized as Map/Reduce tasks via the Hadoop framework. The experiments show the real-time performance for a single robot navigation in two scenarios: the traditional method as sequential algorithm and the presented scheme which is used to estimate fast a parallel algorithm. Also, the contribution of this paper is segmentation of the FastSLAM 2.0 algorithm to execute concurrently the localization and the mapping tasks on the cloud for overcoming strong real-time constraints of the localization task.