To accurately recognize a fault like a bottle neck latent in an entire network by estimating characteristics such as performances and faults of individual parts in the network in accordance with a plurality of observation information of spread end-to-end-paths.
Performances and faults of individual parts (links) in the network are estimated from observation information of end-to-end paths. That is, the throughput of each path spread over the network is measured, and points are given to its distribution. The points are automatically reported to a control tower A from each of end terminals X and Y. The control tower A notices one path spread over the network and determines a maximum value and a minimum value of a section in accordance with the distribution on the basis of the throughput value. Next, the control tower A picks up and considers one link to obtain upper limit values and lower limit values of all the links. The control tower A notices the entire network and evaluates all the links on the basis of a threshold.
MINAMI HIROYOSHI
NOGAMI SHINYA