摘要 |
PROBLEM TO BE SOLVED: To efficiently perform decentralized processing on a series of data to be sequentially input.SOLUTION: A master node 101 in a decentralized processing system 100 successively performs Map processing when detecting an input of a split that a measurement file is divided. For example, when the master node 101 detects an input of a split 3 at a time t3, the master node 101 performs Map processing 3. In the same way, when the master node 101 detects an input of a split 4 at a time t4, the master node 101 performs Map processing 4. At the time t4, the master node 101 determines that tail information 114 is imparted to the split 4. At a time t5, the master node 101 allocates shuffle & sort processing 1 and 2 summarizing the Map processing results 1-4 to a slave node 102#1 and a slave node 102#2, respectively, on the basis of attributes of the Map processing results 1-4. |