Абстрактный

Extract Transform and Load Strategy for Unstructured Data into Data Warehouse Using Map Reduce Paradigm and Big Data Analytics

P.Saravana kumar, M.Athigopal, S.Vetrivel

Analytics over the huge volume of data is now possible with Big data. Data keep on accumulated on every minute from multitude data sources such as social media, mobile devices, and sensors. In order to extract insights from diverse information feeds from multiple, often unrelated sources, data need to be correlated or harmonized to a common level of granularity. Loading Unstructured Data into Data warehouse getting complex. A strategy for fetching the unstructured data into Hadoop Distributed File System is discussed. Data cleansing and profiling of extracted data is important to overcome data quality concerns. Transform phase carried with map reduce frame work. Computation ratio, Network band width and Data locality parameters are monitored with full dump and Incremental load operations. Pig Latin is used to process data from Hadoop Distributed File System and finally load the process data into HDFS file or Data warehouse. Aggregated data from Pig is minimal Subset of Data is Loaded to Data warehouse for Business Analytics and Enterprise Reporting. Based on the Performance related parameters appropriate strategy is suggested for Different type of application.