Our beloved NTP protocol appears to work in a deep space environment
(as tested in a simulation with a 4 hour RTT):
2 Likes
Well I’m no expert at this, but I’m pretty sure the distance wouldn’t be fixed. If the device trying to synchronise time with Earth is travelling, then the response would take a little longer to arrive than the request does, and if we’re on another planet, then I believe they also move in relation to Earth which can change the distance.
Still interesting that NTP works this well with such a long latency, though would be interesting if they would take movement into account, how much that would affect the accuracy
1 Like
There are also relativistic effects that might need to be taken into account. IIRC ntpd was used on some space missions and it has some space-specific code, disabled by default.
I think most NTP clients would ignore measurements with delay larger than 16 seconds, following the RFC, and would need to be patched.
2 Likes
Even for a direct point-to-point linkt, there would need to be some additional fields in the protocol where clients and servers can state their orbital parameters (if they are in an orbit), or their current vector through the solar system. Since the plan is to use relay satellites in orbits around different planets or moons (or the Sun), and to also relay packets using any other spacecraft that picks up a signal, and to dynamically route packets (using store-and-forward where needed), there would also need to be a way for each relay to add its own information, as well as the delay it added between receiving and forwarding a packet. Then times for the different legs of the round-trip could be pretty exactly calculated, allowing for way better accuracy.
An unmodified NTP running over unmodified UDP/IP won’t ever deliver the precision needed for e.g. planning and controlling the engine burn of a spacecraft. The protocols were just not designed to do this.
As far as I know deep space equipment are carrying their own atomic clocks on board. Independently of that, the aforementioned simulation is a nice experiment to find or even expand the limits of the NTP protocol.
On the apropos of talking about the limits of the NTP protocol. It is a bit off topic here, sorry. I propose the following experiment: having a system where the motherboard clock ticks twice the speed of the normal, and another one, where it ticks half speed. How nicely would the different NTP server implementations able to correct such a huge error? Server code, kernel code are permitted to patch before as preparation if there is a hard-coded maximum correctable error limit.
2 Likes
How’s does the “NTP in deep space” experiment account for Einstein’s Time Dilation effect.
P.S. Meanwhile back on earth, we are still dealing with CrowdStrike.
2 Likes
I think so. Even GPS/GNSS satellites have relativistic corrections to account for the altered time they experience being farther from Earth’s center of mass. This time-warping has been experimentally verified with two terrestrial atomic clocks synchronized then one taken atop a mountain for some time then returned and compared with the one that remained at low altitude.
1 Like
Don’t forget the Hafele-Keating experiment flying two atomic clocks around the earth 
link
1 Like