[ntp:questions] Tighter regulation?

unruh unruh at invalid.ca
Sat May 25 14:27:24 UTC 2013


On 2013-05-25, Mischanko, Edward T <Edward.Mischanko at arcelormittal.com> wrote:
>
>> -----Original Message-----
>> From: Charles Swiger [mailto:cswiger at mac.com]
>> Sent: Friday, May 24, 2013 1:01 AM
>> To: Mischanko, Edward T
>> Cc: questions at lists.ntp.org
>> Subject: Re: [ntp:questions] Tighter regulation?
>> 
>> On May 23, 2013, at 10:02 PM, "Mischanko, Edward T"
>> <Edward.Mischanko at arcelormittal.com> wrote:
>> > It takes too long to figure out it needs a more aggressive correction.
>> > If I leave maxpoll at the default of 1024 seconds, my clock drifts
>> outside
>> > of 5 milliseconds consistently.
>> 
> [Mischanko, Edward T] 
>
> Measured by the best time sources I have available.  I am saying that my

And maybe it is that "best" that is wrong. Why not give the details so
we can decide?

>  clock will drift to 5 milliseconds at maxpoll 10 and polling does not
>  decrease to a level that will bring my clock back to an offset of +/- 1
>  millisecond.

Note that on maxpoll 10, the clock will freerun for about 7000 sec
between disciplines. 5ms in 7000 sec is about 1PPM. Now if your computer
heats up and cools down by 20C over an hour that could change the clock
by a few ms. And ntpd is slow to respond to changes in clock rate. 
What happens if you do maxpoll 6? Give us details, not vague statements. 

Also try plotting peerstats and the offset as a function of time. 


>
>> Measured by what?  If you have a better source of time available, sync
>> ntpd using that.
>> 
>> More importantly, ntpd should be entirely able to compensate for a steady
>> drift of ~50 ppm;
>> are you saying that not only do you have a long-term drift, but short-term
>> instability
>> which varies by ~50+ ppm hour-by-hour?
>> 
>> > Too much assumption is made that everyone will have the perfect computer
>> and
>> > the perfect network when configuring these various filters.  What works
>> on
>> > the blackboard does not always work in reality.
>> 
> [Mischanko, Edward T] 
>
> I your 25 year experience have you ever thought they may be a better way to
>  achieve the same goals?  Or, have you ever looked at your goals and asked
>  yourself what are the primary goals and what are secondary goals?  I am
>  suggesting that clock offset is primary to clock jitter.
>
>> Oddly enough, my ~25 year experience with ntpd suggests a great deal of
>> practical
>> experience has gone into creating a timekeeping solution that avoids
>> chasing short-term
>> transients in favor of stable long-term behavior.
>> 
>> However, it's certainly OK if someone decides that some other software
>> provides a better
>> solution for their particular circumstances....
>> 
>> > My computer has a -19 precision but it can't keep time inside 1
>> millisecond with default
>> > Settings; go figure.
>>
> [Mischanko, Edward T] 
>
> I have common PC running Windows XP with 10 peers on my side of a corporate
>  firewall.
>  
>> Precision of -19 suggests commonly available commodity hardware.  Keeping
>> time to around 1ms
>> should be reasonably doable with a decent network connection, at least 4
>> reasonable peers
>> or time sources to query, temperature-controlled systems, and an OS with
>> sane timekeeping;
>> VMs need not apply regardless of OS.
>> 
>> It might help to setup a subnet local peering of ~4 or so machines, in
>> addition to the
>> remote time sources or a GPS/ACTS/WWVB or similar stratum-1 source.
>> 
>> Regards,
>> --
>> -Chuck



More information about the questions mailing list