| 摘要 |
PROBLEM TO BE SOLVED: To provide a technique that enables faster decentralized processing executed under an environment in which there are a plurality of information processing devices in each of which a virtual machine is created.SOLUTION: A Hadoop management unit 14 on a management host 1 comprises a state determination unit 143. The state determination unit 143 determines the processing state of each Hadoop virtual machine 23 from HDFS writing/reading, MapReduce processing, and shuffle and sort processing on the basis of a CPU use rate reported by each host 2, the amount of HDFS communication to access a hard disk device 3 via a DAN 3, and the amount of MapReduce communication performed between Hadoop virtual machines 23. The Hadoop virtual machines 23 whose processing states are determined to be HDFS writing/reading, or shuffle and sort processing are aggregated on one host 2. This allows communication between the Hadoop virtual machines 23 to use more virtual passes in the host 2. |
