[time-nuts] Fwd: NIST UT1 time service

Mike Cook michael.cook at sfr.fr
Tue Jul 21 18:37:21 UTC 2015


Hi,
  Some first tests of the NIST UT1 service recently announced. Forwarded comms with Judah Levine. I’d be interested to hear the experiences of others . 
Regards,
Mike

> Début du message réexpédié :
> 
> De: Mike Cook <michael.cook at sfr.fr>
> Objet: NIST UT1 time service
> Date: 21 juillet 2015 20:23:25 UTC+2
> À: judah at jilau1.colorado.edu
> 
> Judah,
>    First impressions:
>  Yesterday evening I started monitoring the server with ntpd , configuring it with the « no select » directive to prevent ntpd using it as a reference. 
> 
> The other servers that were configured were all from the NTP pool with request transiting the same static IP, which is in fact a VPN link.
> 
> I have attached  a graph of the offsets reported. It covers from 0h00m00s to a couple of minutes ago:
> 
> Some comments
> The first wiggly bit is recorded with the ntp con as described above. The servers sync to UTC at the time , being plus or minus 2ms , which could be related to the very long poll interval I had configured at the time, and the fact that I was running the VPN link. So at around 06:26 UTC I reconfigured the server to use two local GPS sync’d servers. The local offset immediately reduced to micro-seconds. Even though the delay for the UT1 server was around 200ms, jitter was  low, for example from 08:28 UTC:
> 
> Tue Jul 21 10:28:09 CEST 2015
>      remote           refid      st t when poll reach   delay   offset  jitter
> ==============================================================================
> +192.168.1.23    .GPS.            1 u   12   64  377    0.410   -0.066   0.006
>  128.138.140.50  .NIST.           1 u   36   64  377  191.606  303.428   0.799
> *192.168.1.4     .PPS1.           1 u   10   64  377    0.630    0.010   0.014
>   
> As you can see above the reported UTC-UT1 delta is around 303ms and it remained stable for some time. I had a look at the Bulletin A figures and it is not far off for the days predicted value, but is nearer that of Jul26. I’d be interested in knowing what you see locally. 
> 
> Around 11:48 UTC the jitter became high :
> 
> Tue Jul 21 13:48:13 CEST 2015
>      remote           refid      st t when poll reach   delay   offset  jitter
> ==============================================================================
> +192.168.1.23    .GPS.            1 u   62   64  377    0.420   -0.109   0.021
>  128.138.140.50  .NIST.           1 u   48   64  377  201.419  298.414   3.392
> *192.168.1.4     .PPS1.           1 u   11   64  377    0.645   -0.040   0.018
> 
> I was doing some admin on that system (modifying a python app to send NTP requests to take ntpd out of the loop) which required some file transfers. Anyhow, as you can see, the reported offset dropped dramatically and on the graph you can see that it has stayed unnaturally low.  That said, factoring in the jitter,delay we still have a ballpark figure close to the stable value. However, not ideal. 
> 
> Your server is a long way from me. 155ms via a non VPN link and 200ms over VPN. That shouldn’t affect things much unless I am getting an asymmetric path, which is possible. 
> 
> Next test will be to so how the offset appears using a non ntpd polling. I’ll let you know how I get on.
> 
> Regards,
> Mike
> 
> PS. If my images don’t pass your firewall, you can check the cubieez2 peerstats link on <http://stratum1.ddns.net:8080 <http://stratum1.ddns.net:8080/> >
> 
> 
> 
>> Le 20 juil. 2015 à 23:47, Judah Levine <judah.levine at colorado.edu <mailto:judah.levine at colorado.edu>> a écrit :
>> 
>> Hello,
>>   Yes, I would be interested to learn what you find. My rough estimate is that the accuracy of NTP is typically not better than about 5% of the round-trip delay.
>> 
>> Judah Levine
>> 
>> 
>> On 7/20/2015 3:27 PM, Mike Cook wrote:
>>> Thanks Judah,
>>>   
>>> 
>>>> Le 20 juil. 2015 à 17:25, Judah Levine <judah.levine at colorado.edu <mailto:judah.levine at colorado.edu>> a écrit :
>>>> 
>>>> 
>>>> Hello,
>>>>   Your address is registered and should be ready to go. I am sure that you understand that the accuracy of the time messages will depend on the stability of the network connection from your system back to Boulder.
>>> Yes, that is understood. If you are interested in what I see as a client in my neck of the woods and any comments I may have, I will update you.
>>> 
>>>>    I have a number of ideas about the second point, but I have no requests so far, and I am waiting to see if anybody actually requests that service. There are a number of questions about the details that I would discuss with real users.
>>>  Understandable.
>>> 
>>> Regards,
>>> Mike
>>> 
>>>> Best wishes,
>>>> 
>>>> Judah Levine
>>>> Time and Frequency Division
>>>> NIST Boulder
>>>> 
>>>> 
>>>> On 7/20/2015 2:09 AM, Mike Cook wrote:
>>>>> Hi Judah,
>>>>>       Could you please register IP 77.78.103.22 for me.  Thanks
>>>>> I don’t have domain name registered with it at the moment.
>>>>> reverse DNS returns
>>>>> 
>>>>> 22.103.78.77.in-addr.arpa    name = assigned-77-78-103-22.dc3.cz.
>>>>> 
>>>>> On the second proposed service described in <http://www.nist.gov/pml/div688/grp40/ut1_ntp_description.cfm <http://www.nist.gov/pml/div688/grp40/ut1_ntp_description.cfm>> I would just mention a couple of approaches that you may be aware of already and discussed at length by contributors to the leapsecs at leapsecs.com <mailto:leapsecs at leapsecs.com> mailing list with test implementations. The threads are somewhat dispersed however.
>>>>> 
>>> "Ceux qui sont prêts à abandonner une liberté essentielle pour obtenir une petite et provisoire sécurité, ne méritent ni liberté ni sécurité."
>>> Benjimin Franklin
>> 
>> -- 
>> Judah Levine
>> JILA
>> University of Colorado
>> Boulder
> 
> "Ceux qui sont prêts à abandonner une liberté essentielle pour obtenir une petite et provisoire sécurité, ne méritent ni liberté ni sécurité."
> Benjimin Franklin
> 

"Ceux qui sont prêts à abandonner une liberté essentielle pour obtenir une petite et provisoire sécurité, ne méritent ni liberté ni sécurité."
Benjimin Franklin




More information about the Time-nuts_lists.febo.com mailing list