DISTRIBUTED PROCESSING APPARATUS AND METHOD FOR BIG DATA USING HARDWARE ACCELERATION

摘要

PURPOSE: A distribution processing device of massive data and a method thereof are provided to improve the dispersion parallel processing speed of data by executing a parallel process in data inputted through a sub-unit. CONSTITUTION: A mapping node (120) includes one or more mapping units and outputs a mapping result by executing a parallel process in an inputted data. A shuffle node (140) includes a shuffle unit and a memory buffer and records a shuffle result in the memory buffer by processing the mapping result in parallel. A combination node (150) includes a combination unit and generates a combination result by combining the content of the memory buffer through the combination unit. An input distribution node (110) distributes the inputted data to the mapping units. [Reference numerals] (110) Input distribution node; (120) Mapping node; (130) Output distribution node; (140) Shuffle node; (150) Combination node; (160) Combining node; (170) Refining node