[ntp:questions] Re: NTP precision

Tom Smith smith at cag.zko.hp.com
Sun Sep 25 23:55:41 UTC 2005


Tom Smith wrote:
> Brad Knowles wrote:
> 
>> At 4:03 PM +0000 2005-09-23, Tom Smith wrote:
>>
>>>  However, I'm not sure of the proper way to put these together.
>>
>>     We've tried to put some useful definitions at 
>> <http://ntp.isc.org/bin/view/Support/NTPRelatedDefinitions>.  This may 
>> be of some help to you.
>>
>>     If you have any further questions on these definitions, or you 
>> have any suggestions for clarification, you can either make those 
>> yourself via the standard TWiki procedure, or you can mention them 
>> here and we will try to make the appropriate updates at some point in 
>> the hopefully-not-too-distant future.
>>
> 
> Thanks Brad.
> 
>  Jitter
>     Short-term variations in Frequency with components greater than 10 Hz.
>     The estimated time error of the system clock measured as an exponential
>     average of RMS time differences. [This needs some work. :-)]
> 
>  Dispersion
>     Represents the maximum error of the local clock relative to the
>     reference clock. [reported only for individual peers using "rl [ID]"]
> 
>  Root Dispersion
>     This is a number indicating the maximum error relative to the primary
>     reference source at the root of the synchronization subnet, in seconds.
>     Only positive values greater than zero are possible. [It's in 
> milliseconds
>     in ntpq]
> 
> Reading between the lines, and adjusting for the actual units reported
> by "ntpq -c rl 0" (milliseconds), it seems to say:
> 
> Estimated error of the current local system time[?] relative to [what?]:
> 
>     "true" time[?] = systemtime[?] +- jitter
> 
> Maximum error of the current local system time[?] relative to the source
> at the root of the currently selected chain of sources:
> 
>     root server time = systemtime[?] +- rootdispersion
> 
> Maximum error of the current local system time relative to
> "true" time:
> 
> [???]    "true" time = systemtime +- rootdispersion [+- some other error?]
> 
> It's still not crystal clear to me, so I don't really know how
> to improve the wording.
> 
> -Tom

While the discussion of the real meaning of "jitter" goes on elsewhere,
Leandro still has the problem of how to express the uncertainty of the
current system time. From the data returned by "ntpq -c rl 0", I'm
coming to the conclusion that:

    "true" time = systemtime + offset +- (rootdispersion + some other error)

Where "some other error" may or may not be the "jitter" of the root
server (the stratum 0 server/reference clock), and where "true time"
would have to be interpreted as meaning the "true time as perceived
by the root server" - probably very small compared to the other
error components. The "offset" is at any time the remaining difference
to be worked off to agree with the consensus time among all the peers,
whereas the "rootdispersion" includes the aggregate contribution of delays
only through the currently selected peer chain back to its root. Not quite
apples and apples, but that seems to be about as close as you can get.
Does that seem right?

-Tom




More information about the questions mailing list