[ntp:questions] Help appreciated

Richard B. Gilbert rgilbert88 at comcast.net
Wed May 13 21:44:32 UTC 2009


Towli wrote:
> 
> "David Woolley" <david at ex.djwhome.demon.co.uk.invalid> skrev i 
> meddelelsen news:4a0b04b5$0$516$5a6aecb4 at news.aaisp.net.uk...
>> Towli wrote:
>>>
>>> "David Woolley" <david at ex.djwhome.demon.co.uk.invalid> skrev i 
>>> meddelelsen news:4a0ab078$0$513$5a6aecb4 at news.aaisp.net.uk...
>>
>>>>
>>>>>> #sh ntp ass de
>>>>
>>>> What command are you running?
>>>
>>> Show ntp association detail  - sry
>>
>> What OS are you using; that command is not part of the standard ntpd 
>> distribution.
> 
> Im using Cisco 3750 switches synching with a DFC77 appliance box and 
> external (InetProviders) Stratum 1 servers.
> 
> I am really trying to understand this, and i feel a little stupid, 
> -perhaps i dont get it because english is not my native language:
> Why is my 3750 switch reporting a precison of 2**-6 when using my 
> appliance DCF77 box, but reporting precison of 2**-18 when talking about 
> the external Gps-servers.
> If i erase the ntp clock-period from my switches when they are in synch 
> with my appliance box - would the precision improve ?
> 
> /T
> 
> 

Precision is the smallest possible difference between two successive 
readings of the clock.  It has nothing to do with accuracy.  Precision 
is usually specified as a negative power of two; e.g. a precision of -10 
says that the clock "ticks" each represent 1/1024 seconds or 
approximately one millisecond.  Solaris, running on SPARC hardware has a 
precision of -21 which is somewhere near 1 microsecond.

Your switches were designed to switch packets!  Any timekeeping they do 
is almost certainly for the purpose of time stamping log files.  The 
events they might timestamp probably occur on a scale of milliseconds, 
centiseconds, or even deciseconds (1/10 seconds).

And don't worry about your English; it's certainly better than my Danish 
(non-existent) or my Deutch (almost non-existent).




More information about the questions mailing list