Windows-8 with Internet servers alone
Data from PC Torvik, with no LAN servers. This PC does
seem to have some challenges to good timekeeping introduced, perhaps, by the
BIOS. In the results below, this may be the cause of the frequency change
from around 02:00 UTC.
Windows-8 with LAN servers
Data from a different PC, a portable PC running the release version of
Windows-8. This is a better timekeeper in any case, and here it's syncing
to stratum-1 NTP servers (prefer is a FreeBSD server) on the local
network. The minpoll and maxpoll are both set to 5 (32 seconds polling
interval). The Internet servers are still present, but set to a fixed long
poll interval (1024 s, minpoll =10) Here we see much better performance, with jitter averaged
jitter between 40 and 80 µs, and offsets between -220 and +320
µs. The performance is even more surprising as this PC is connected
via a Wi-Fi link to the main LAN.
Here are some snapshot ntpq results. The jitter
on two of the LAN servers seems very high, but the reported ntp jitter is low as
can be seen from the graphs above. At least some of the high jitter is
from the Wi-Fi connection, but not all of it, I suspect.
C:\>ntpq -p bergen
remote refid st t when poll reach delay offset jitter
==============================================================================
*pixie .PPS. 1 u 26 32 377 1.301 0.089 43.116
+feenix .PPS. 1 u 2 32 377 1.301 0.083 0.417
+stamsund .PPS. 1 u 3 32 377 2.035 0.526 77.736
uk.pool.ntp.org .POOL. 16 p - 1024 0 0.000 0.000 0.002
-linnaeus.inf.ed 129.215.64.32 2 u 28 1024 377 32.605 3.957 6.514
-time.mhd.uk.as4 213.229.82.130 3 u 704 1024 377 23.502 11.060 4.196
-dns0.rmplc.co.u 193.62.22.74 2 u 515 1024 377 23.974 3.783 38.570
-ns2.c-it.co 195.66.241.2 2 u 317 1024 377 22.255 3.640 11.516
-ns1.c-it.co 195.66.241.3 2 u 534 1024 377 23.874 4.562 4.750
-protium.cloudfa 80.83.126.67 3 u 547 1024 377 25.710 4.612 1.158
C:\>ntpq -p bergen
remote refid st t when poll reach delay offset jitter
==============================================================================
*pixie .PPS. 1 u 6 32 377 1.301 0.089 0.370
+feenix .PPS. 1 u 14 32 377 1.301 0.083 0.417
+stamsund .PPS. 1 u 15 32 377 2.035 0.526 77.736
uk.pool.ntp.org .POOL. 16 p - 1024 0 0.000 0.000 0.002
-linnaeus.inf.ed 129.215.64.32 2 u 40 1024 377 32.605 3.957 6.514
-time.mhd.uk.as4 213.229.82.130 3 u 716 1024 377 23.502 11.060 4.196
-dns0.rmplc.co.u 193.62.22.74 2 u 527 1024 377 23.974 3.783 38.570
-ns2.c-it.co 195.66.241.2 2 u 329 1024 377 22.255 3.640 11.516
-ns1.c-it.co 195.66.241.3 2 u 546 1024 377 23.874 4.562 4.750
-protium.cloudfa 80.83.126.67 3 u 559 1024 377 25.710 4.612 1.158
Windows-7 with Internet servers alone
Data from PC Ystad, showing LAN servers (until 15:38 UTC Oct
03), Internet servers alone until 19:30 Oct 04, and then back to LAN
servers. One LAN server (192.168.0.3) was left for purely monitoring
purposes in the peerstats, it was configured with NOSELECT to ensure that it no
part in disciplining the NTP clock.
Here are the ntpq -p results after running for one day with
just Internet servers.
C:\>ntpq -p ystad
remote refid st t when poll reach delay offset jitter
==============================================================================
uk.pool.ntp.org .POOL. 16 p - 64 0 0.000 0.000 0.977
-ntp.uk.gbit.ws 192.93.2.20 2 u 571 1024 377 40.948 12.089 10.964
-95.172.20.162 ( 129.215.160.240 3 u 378 1024 377 22.956 -4.388 9.264
-b0ff54d9.bb.sky 131.188.3.221 2 u 342 1024 377 44.910 7.057 4.801
-panzer.charliec 245.200.202.182 4 u 332 1024 377 28.999 58.721 31.026
-78.33.40.170 (7 81.2.117.235 2 u 526 1024 377 34.990 9.832 10.149
+dns0.rmplc.co.u 195.66.241.2 2 u 350 1024 377 24.818 0.234 10.879
*ntp.thirdlight. 193.67.79.202 2 u 269 1024 377 24.954 0.769 9.807
+resntp-b-vip.lo 20.139.208.232 4 u 203 1024 377 32.894 6.338 6.968
+indium.cloudfab 195.66.241.2 2 u 61 1024 377 25.918 0.355 12.525
| 
