To make processing time equal for each node as much as possible by deciding the size of data to be divided based on state information, dividing the data and distributing them to respective computers consisting of a distributed system.

A control software 2700 is composed of a kernel 3100 and an inter-node pipe control part 3150. A kernel 311 performs the management, etc., of resources provided for computers. The inter-node pipe control part 3150 is composed of a state information monitor part 3200, data size deciding part 3250, allocate computer deciding part 3300 and pipe control part 3500. Then, the state information of respective computers consisting of the distributed system is possessed, the size of data to be divided is decided based on the state information, and data are divided and distributed to the respective computers consisting of the distributed system. In this case, the computers consisting of the distributed system are divided into groups, any group to which the computer to distribute data belongs is designated, the number of data to be divided is designated, and the computers are selected as many as the number of divided data.