The present invention includes the steps of receiving the length of the record (A) of the binary data; (B) Hadoop distributed file system, the closest point to the starting block of n is a multiple of the length of the record from the data block must be processed as a starting point of the block of data stored in (HDFS) and setting the previous InputSplit boundary of their InputSplit defining a InputSplit by; (C) generating a RecordReader and returns it to perform work by the length of the record to read from the starting point for their entire area InpuSplit defined above; And (D) a step of extracting said record in the form of a (Key, Value) through RecordReader (LongWritable, BytesWritable); Hadoop for processing the binary data with the distribution of the fixed-length records, characterized in that comprises a input format and in MapReduce, to an analysis method for binary data using the input format. According to the input format of the present invention, since the binary data of a fixed length to be processed in a distributed Hadoop environment without changing the data format operation processing is possible, requiring less storage space compared to other types of data and enables faster processing speed The. ;
展开▼
