Research And Implementation Of A Mapreduce Computation On A Reconfigurable Hardware Architecture

With the rapid development of the Internet and the explosive growth of information, efficient scientific computing is required to meet the needs of quick valuable information mining. It takes a long time and great energy to process massive data on the traditional computer. How to quickly and efficiently make use of the mass data is a severe test for all industries and, following this, it is an inevitable choice for the development of the industry to use massive data for valuable information mining.MapReduce programming model is proposed by Google Labs. It is a fast, simple and efficient method for processing and generating large data sets. Programs written in this functional style are automatically parallelized and executed on a large cluster of commodity machines. This allows programmers without any experience with parallel and distributed systems to easily utilize the resources of a large distributed system. By comparison, the performance and power consumption of the traditional computer, with its limited unit and memory bandwidth, can not meet the increasing demands of computing in application. Different from this, the processing system with reconfigurable computing architecture combines the advantages of efficiency of application specific integrated circuit (ASIC) and flexibility of general-purpose processor.This paper conducts a deep research into the MapReduce programming model and reconfigurable computing technology in terms of efficient processing of massive data, and provides an approach that combines the MapReduce programming model with reconfigurable computing technology. To complete the mapping of these two algorithms on reconfigurable hardware architecture, Kmeans algorithm and FIR algorithm, two typical algorithms, are selected as the target algorithms and the programming of Kmeans algorithm and FIR algorithm is also designed. Eventually, Kmeans clustering algorithm and FIR filtering algorithm are mapped on the reconfigurable computing architecture. By comparison, Kmeans clustering algorithm and FIR filtering algorithm are also executed on the general purpose computing platform. Results show that on the basis of the proposed scheme, the performance of time is improved 3.2 times and 2.6 times compared with the general-purpose computing platform.