[time-nuts] NTP Time Server Ping Timing Stability

[Hal Murray]

> I'm curious about the stability of Ping timing between high quality
> NTP  servers.  Is it the case that a histogram will show different
> times because the  path may be different between the two servers but
> the spread for any given path  would be quite small because of the
> quality of the timing? 

NTP measures the transit time in both directions.  It assumes symmetry and 
calculates the offset as half the difference and uses that to adjust the 
local clock.

Assuming you have good clocks on both ends:

If you plot the offset vs round trip time, you get an arrow pointing left.  
The point (blob) of the arrow is the best case.  The edges of the arrow are 
due to queuing delays in one direction or the other.  You also get some 
scattered points trying to fill in the arrow from queuing delays in both 
directions.

If you plot round-trip time vs time of day, you can see occasional jumps 
where the network path changes, or somebody starts a big download that fills 
up one side of a pipe for a while.


> Or if I was pinged would the wander of the satellite in the link show
> up or is  my stock PC so crude it swamps the satellite wander?
> 148.78.32.98 or 148.78.32.97 not sure if those are visible on the net.

ping just gives you the round trip time.  I'm not sure where ntp fits in.

It's reasonable to get ntp with a GPS clock running in the few microsecond 
range.  I'd guess you might see satellite wander if that covers a few miles.  
It might be hard to see in all the other network noise.

Where are you located (packet hops, not crow fiies) relative to the satellite 
ground stations?







-- 
These are my opinions, not necessarily my employer's.  I hate spam.