>>> I've seen so far - 1080 requests during one single second. Such
>>> extrema occur sometimes but always precisely at a full hour, probably
>>> due to a large number of ntpdate or sntp users with their periodic jobs.

> What is the averaging time for your rps values? Sometimes you can get
> say 6000 requests per minute although 1500 of them arrived at one single
> second. For such a case even xx/1 Mbps line is slow, any packets get an
> asymmetric delay. Of course, this is not a problem for any ntpd client
> with a good configuration.

I would say that this isn't a problem.  Assuming the spikes when there
would be the asymmetric delay are mostly sntp and ntpdate users (where
high accuracy isn't as much of an issue) the error introduced by the
delays aren't much of a problem.  The actual ntp clients will compensate
for this anomalous behavior with the continuing polls throughout the
rest of the hour.  Correct?

sjm
_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers

Reply via email to