If I've understood the original question correctly then this monitoring
should be done from the server. So there are several possibilities:
1.) run ntpq -p on the server but actually send the request to the client:
ntpq -p <client-ip>
2.) if the client does not respond to 'ntpq -p' requests, use ntpdate:
ntpdate -q <client-ip>
3.) if monitoring should be done continuously and you have only a couple of
known clients, you can add them to your server's ntp.conf file and tell the
server just to poll them but not to try to synchronize to the clients.
server <client-ip> noselect
In the latter case, if you run 'ntpq -p' on the server you also get a list
which tells you the estimated time offset of each client.