>> Best round trip is almost the same as the IPv4 path (186.3ms), but the
>> average is 194.7 and the slowest is 201.8ms.  Much harder to get an accurate
>> time if each packet can take that much different time.

> It would be important to clarify if this is a general effect or
> affects LA monitor point only.
> In former case NTP over IPv6 should not be recommended in general.

A difference in +/- 15ms out of 200ms round trip time, just is not that big a 
deal. That's sort of jitter is entirely characteristic of most high-latency 
network paths and is far from worst-case.

ntpd closely tracks the round trip time and jitter in round trip time between 
all sources and uses a very wise algorithm to reject any that don't look good. 
If a source with 200ms latency and 15 ms jitter is rejected, it's probably 
because of its latency, and not because of its jitter.

ntpd is also very good at excluding individual samples with high popcorn 
spikes. These drive the pool scores low rapidly yet ntpd has no problem at all 
dealing with time sources with this kind of random network asymmetry.

Tim.

_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool

Reply via email to