On 2020-01-10 02:55 PM, Dave Howorth wrote:
Once upon a time I worked with a telco setting up an ATM network in London. Their time source was in a rack with five slots. Slot 1 was a GPS card. Slot 2 (backup) was a GPS card. Slot 3 (backup's backup etc) was a GPS card. Slot 4 was a GPS card. Slot 5 was an atomic clock.
Many years ago, I had something to do with a clock for the old "T1" TDM timing, which used the Loran C signal to provide the time base. However, that was for the signal clock, not time of day. In my work, a few months ago, I saw a requirement for NTP that was clearly written by someone who didn't understand it. This was for a transit system, where a PPP (Public Private Partnership) is building a new line. The requirement, from the transit company said the PPP company was to use the existing NTP server, but if it failed, they were to switch over their (PPP) server. Apparently, whoever wrote that didn't understand peering, multiple sources, etc. or that redundancy and fallback were all built into NTP. The PPP network has 2 GPS receivers which, at stratum 0, would already be more precise than the stratum 1 coming from the transit company. Also, since NTP is supposed to be traceable back to International Atomic Time, there should be no discrepancy between the stratum 0 sources. Then they have to devise some means to detect failures etc.. As I said, someone doesn't know what they're talking about. I'd expect the proper configuration would be to peer the 2 PPP stratum 1 servers and client/server from the transit company stratum 1 servers to PPP stratum 1. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org