A computer-implemented method of carrying out fault diagnosis of a monitored device in an IT network. The IT network has a monitoring station arranged for monitoring the monitored device. A routing path between the monitoring station and the monitored device is traced and stored. In response to an apparent fault of the monitored device at least one request to a device or devices along the stored routing path is sent. The fault is located to within a subnetwork of the IT network based on whether or what sort of a reply to the request is received from the requested device or devices.