A comparison study among GPU and map reduce approach for searching operation on index file in database query processing

Abstract

As amount of data in different form is increased day by day; it is very difficult to process it. This unstructured form of data cannot be easily retrieved through query processing. Normally SQL query acts on structured data. To convert unstructured data into structured data, Hadoop provides map reduce approach. Instead using map function, we can use GPU approach for processing data in parallel and then we can use reduce function on processed data. Here we compare two approach i.e. map-reduce approach and gpu-reduce approach to calculate performance measurement for searching index file. As Hadoop is a framework which is purely based on Java, we use JCUDA programming language to implement gpu-reduce approach.