The offset is being taken at 15 second intervals via a script that calls 
"chronyc -c sourcestats” and feeds the data to rrd. My polling was not quite as 
aggressive, at 2 for the device shown in the graphs I sent.

I wouldn’t think I have much in the way of asymmetry, but anything’s possible. 
Both the monitoring system and the and the NTP servers under test are connected 
to the same switch. Here’s what my sources/sourcestats looked like:

210 Number of sources = 7
MS Name/IP address         Stratum Poll Reach LastRx Last sample               
===============================================================================
^+ 192.168.230.240               1   2   377     3   -218ns[ -240ns] +/-   23us
^x 192.168.230.241               1   2   377     0    -40us[  -40us] +/- 2930us
^x 192.168.230.242               1   2   377     0    -40us[  -40us] +/- 2930us
^* 192.168.230.244               1   2   377     0   -146ns[ -168ns] +/-   23us
=x 192.168.230.1                 2   3   377     1  -7027ns[-7049ns] +/- 1530us
=x 192.168.230.2                 2   2   377     4  +4821ns[+4799ns] +/-   86us
=x 192.168.225.10                2   2   377     5    -24us[  -24us] +/-  206us
210 Number of sources = 7
Name/IP Address            NP  NR  Span  Frequency  Freq Skew  Offset  Std Dev
==============================================================================
192.168.230.240            47  28   202     -0.001      0.005    +53ns   558ns
192.168.230.241            25  14    97     -0.015      0.208    -41us  7733ns
192.168.230.242            16   6    60     -0.162      0.355    -48us  5623ns
192.168.230.244            26  13   105     +0.002      0.010    -52ns   410ns
192.168.230.1              39  17   355     -0.002      0.023   +361ns  3795ns
192.168.230.2              48  24   251     -0.008      0.020   +382ns  2505ns
192.168.225.10             64  38   606     +0.003      0.013    -17us  5355ns

23/24us for the error estimate isn’t bad considering that the minimum round 
trip transmission time through the switch is 21.44 us. If I direct connect the 
server (bypassing the switch), then this drops down to 21/22us (the switch is 
3.44us).

Back on the switch, if I bring the polling down to 1 then things tighten up a 
bit more. But still nowhere near the 151ns standard deviation that you are 
getting.

210 Number of sources = 7
MS Name/IP address         Stratum Poll Reach LastRx Last sample               
===============================================================================
^+ 192.168.230.240               1   1   377     0   -212ns[ -212ns] +/-   24us
^x 192.168.230.241               1   2   377     0    -27us[  -27us] +/- 2930us
^x 192.168.230.242               1   2   377     1    -43us[  -43us] +/- 2930us
^* 192.168.230.244               1   1   377     1   +174ns[ +157ns] +/-   24us
=x 192.168.230.1                 2   3   377     7  +2809ns[+2835ns] +/- 1552us
=x 192.168.230.2                 2   1   377     3  -1394ns[-1411ns] +/-   82us
=x 192.168.225.10                2   1   335     0    -31us[  -31us] +/-  191us
210 Number of sources = 7
Name/IP Address            NP  NR  Span  Frequency  Freq Skew  Offset  Std Dev
==============================================================================
192.168.230.240            16   6    30     +0.005      0.029     +4ns   264ns
192.168.230.241            64  28   255     +0.029      0.038    -25us  6948ns
192.168.230.242            39  23   154     +0.005      0.102    -46us  7070ns
192.168.230.244            45  27    92     -0.001      0.005    -23ns   237ns
192.168.230.1              23  13   210     +0.015      0.033  +1262ns  2521ns
192.168.230.2              64  29   215     +0.007      0.012  -2301ns  1733ns
192.168.225.10             63  33   474     +0.003      0.019    -19us  5579ns

Denny



> On Nov 03, 2016, at 23:51, Miroslav Lichvar <mlich...@redhat.com> wrote:
> 
> On Thu, Nov 03, 2016 at 01:38:25PM -0700, Denny Page wrote:
>> I’m a little late to the party, but I thought it would be worth visually 
>> noting the positive impact that the nanosecond timestamp changes have had. 
>> The first graph is from a dedicated chrony monitoring system. The transition 
>> from 2.4 to the current version occurs midway though the graph. Nothing too 
>> dramatic, but still a noticeable improvement. The second graph is from a 
>> more active general server. The improvement here is much more significant.
> 
> Which offset do the graphs show? The one from statistics.log? The
> improvement might be related also to the correction for asymmetric
> jitter that was included recently.
> 
>> Thank you for all the hard work. I’m really looking forward to hardware 
>> timestamps. :)
> 
> I'm busy with other things right now, but I hope I'll have something
> ready for testing soon. It's a lot of new code and I need to put it in
> a better shape before I can push it to git. Anyway, here is how my
> chronyc sources and sourcestats look right now :).
> 
> 210 Number of sources = 1
> MS Name/IP address         Stratum Poll Reach LastRx Last sample              
>  
> ===============================================================================
> ^* 192.168.100.1                 1   1   377     1   +259ns[ +267ns] +/-   
> 23us
> 210 Number of sources = 1
> Name/IP Address            NP  NR  Span  Frequency  Freq Skew  Offset  Std Dev
> ==============================================================================
> 192.168.100.1              24  11    46     +0.000      0.009     +0ns   151ns
> 
> -- 
> Miroslav Lichvar

Reply via email to