[ntp:questions] Sine wave offset patterns
unruh at invalid.ca
Tue Sep 10 18:17:58 UTC 2013
On 2013-09-10, E-Mail Sent to this address will be added to the BlackLists <Null at BlackList.Anitech-Systems.invalid> wrote:
> Jared Watkins wrote:
>> I've had a chance to try a few things since I posted this.
>> I don't think it's temperature.
>> I have plots of that and it's very stable.
>> I think it's a combination of network latency to the
>> upstream peer combined with varying system jitter.
>> I plotted the jitter and saw that it's a sawtooth pattern
>> that varies over the same interval with peaks and valleys
>> lining up with the offset swings.
>> On 'good' hosts it fluctuates between about .5 and a max of 1ms
>> and on bad hosts it's more like 1-4ms of jitter.
>> A really good host had jitter swings of between about .2 to .5 ms
>> and a corresponding offset swing of about +- 8ms.
>> I don't know for certain this is cause and effect
>> but I've not been able to find anything else.
> Unless you decreased the maxpoll,
> I would have thought the eventual long poll times would
> have smoothed that out.
"long polls(10)" are 20 min. Even with the 1 in 8 of the clock filter,
that is still just over 2 hours.
> The only time I've seen anything like that (twice, in 2 different states),
> was on a overloaded (grossly over sold bandwidth) cable TV ISP network,
> you could watch the latencies (and packet loss) climb,
> as kids got out of school, then later as others got home from work;
> I would have thought ISPs would have improved their
> infrastructure beyond that by now.
Why would they. The fact that he is seeing worse behaviour on distant
sources makes one suspect that it is a "one way" latency issue. There is
no reason why the machine's own behaviour should depend on how far away
the source is.
More information about the questions