CUDA architecture in zero-latency data warehouse
Marcin Gorawski, Damian Lis · 2011
There is a growing importance and the need to ensure data actualisation and efficiency of their processing in zero-latency data warehouses. The standard approach, based on the traditional process of data extraction (ETL) was not sufficient because of the need for time-sharing access during the process of refreshing the data and the time of analysis. The paper presents an ETL system for the zero-latency data warehouse. This system implements the WINE-HYBRIS algorithm based on CUDA and CPU architectures. Presented performance testing of the system, is based on two completely different architectures, allowing the possibility of observing upcoming opportunities that arises during usage of the CUDA architecture in data warehousing systems.