Abstract
Driven by the ever-growing of data from social network (SN), data warehouse (DW) approaches must be adapted. Generally the star, snowflake or constellation models are used as logical ones. All these models are inadequate when dealing with social data which need scalable and flexible systems. As an alternative, NoSQL systems begin to grow.
In the absence of a clear approach which allows the implementation of data warehouse under NoSQL model, we propose in this paper, new rules for transforming a multidimensional conceptual model into two NoSQL ones: column-oriented and document-oriented models. For each model, we distinguish two types of transformation: simple and hierarchical. To validate our transformation rules, we implemented four data warehouses using Cassandra as a column-oriented NoSQL system and MongoDB as document-oriented NoSQL system. These systems were implemented using java routines in Talend Data Integration tool and evaluated in terms of "Write Request Latency" and "Read request Latency" using TPC-DS benchmark. (C) 2016 The Authors. Published by Elsevier B.V.