Disclosed are: a system to dispersedly store a video file in a Hadoop distributed file system (HDFS); a system to perform a MapReduce for a video stored in the HDFS; and a providing method thereof. According to an embodiment of the present invention, the system to dispersedly store a video file in an HDFS includes: a division module which divides the video file into the N number of storage file blocks (wherein N is the number of blocks when the video file is divided into a predetermined default block size); and a control module which dispersedly stores the N number of storage file blocks in a data node of the HDFS. An i^th storage file block (1=i=N) of the N number of storage file blocks includes data included in an i^th basic size file block, wherein the i^th basic size file block is an i^th block when dividing the video file into basic block size units. A j^th storage file block (1=j=N-1) of the N number of storage file blocks further includes data immediately prior to an initial key frame included in a (j+1)^th basic size file block, when the starting point of the (j+1)^th basic size file block is not the starting point of the key frame.
展开▼
