OLAP (online analytic processing) systems provide valuable insights into organizations; thus, it becomes necessary to integrate legacy OLAP systems into scalable and distributable architectures. This project comprises two important tasks: the first is transferring OLAP cubes to share nothing architectures. The second task is integrating OLAP information with other OLAP systems over distributable and scalable architectures. The main problem is to convert conceptual model OLAP data sources to shared nothing architectures. An additional problem is query execution time on the shared nothing architectures because by default, complete data locality is not considered in these environments. In this paper, Arvand is proposed. This method can transfer multidimensional data sources into shared nothing architectures. Data are captured from multidimensional data sources and converted into a unified format. Through unification, multidimensional data sources can be easily distributed over homogeneous and heterogeneous nodes because the nodes will not need additional information from other nodes. As an added benefit, MapReduce methods can be used properly and with maximum performance for query retrieval. Arvand is implemented by adding some components to Hadoop. In this paper, architectures with different heterogeneous and homogenous nodes are proposed and evaluated using a TPC-DS benchmark.