P-ETL: Parallel-ETL based on the MapReduce paradigm

Big data is an opportunity in the emergence of novel business applications such as “Big Data Analytics” (BDA). However, these data with non-traditional volumes create a real problem given the capacity constraints of traditional systems. The aim of this paper is to deal with the impact of big data in a decision-support environment and more particularly in the data integration phase. In this context, we developed a platform, called P-ETL (Parallel-ETL) for extracting (E), transforming (T) and loading (L) very large data in a data warehouse (DW). To cope with very large data, ETL processes under our P-ETL platform run on a cluster of computers in parallel way with MapReduce paradigm. The conducted experiment shows mainly that increasing tasks dealing with large data speeds-up the ETL process.