A physical machine calculates, for each virtual switch, an increment of a usage of a processor in accordance with the number of packets to be collected, by using a usage of the processor by a virtual switch and a number of packets passing through each port when mirroring is executed and when mirroring is not executed, respectively. For each of candidates for capture point, the physical machine calculates, for each physical machine, a usage of the processor by a hypervisor when mirroring is executed, by using an increment of a usage of the processor by the virtual switch and a usage of the processor by the hypervisor when mirroring is not executed. The physical machine determines a combination of ports used for collection of packets from among the candidates for capture point, in accordance with a deviation of the calculated usage of the processor by each hypervisor.