[ntp:questions] Keeping NTP Honest Redux

David J Taylor david-taylor at blueyonder.not-this-bit.nor-this.co.uk.invalid
Mon Nov 9 07:24:43 UTC 2009


"Unruh" <unruh-spam at physics.ubc.ca> wrote in message 
news:6mKJm.51168$Db2.25058 at edtnps83...
> "David J Taylor" 
> <david-taylor at blueyonder.not-this-bit.nor-this.co.uk.invalid> writes:
>
>>"Hal Murray" <> wrote in message
>>news:v72dnRRFBauF8WvXnZ2dnUVZ_t5i4p2d at megapath.net...
>>>
>>>>NTPD is at its best from about 2300 local time to 0700 local time. The
>>>>net quiets down and NTP packets travel with minimal and highly
>>>>predictable delays.
>>>
>>> Except at 3 AM when the cron jobs go off and warm up the system.
>
>>Many such temperature variations must have a fairly strong 24-hour 
>>cyclic
>>component, so would there be an improvement in performance if NTP could
>>remember the frequency (or whatever) averaged over, say, the last seven
>>days, and use this as an expected base value?
>
> except or course weekends, and that Friday when siddenly that damn 500 
> page
> report is due and you have written 2 pages so far.
> Far better for ntp to have measurements of the internal temp and use 
> those to
> compensate.

Yes, Bill, I have some very simple measurements here:

  http://www.satsignal.eu/mrtg/performance_narvik.php

where you can see NTP offset (with a LAN source) plotted next to hard disk 
temperature (the easiest thing I could measure).  In the absence of an 
accurate measure of temperature (in the right area of the system), I still 
think a history (or two - weekday and weekend) would be a worthwhile 
addition.

Cheers,
David 




More information about the questions mailing list