[ntp:questions] NTP/clock behaviour/config

Patrik Arlos Patrik.Arlos at bth.se
Wed Aug 31 10:43:07 UTC 2005


Hi,

 

I've been looking at the stability and accuracy that is obtained for a host
system clock. I've scanned the jun-aug archives and saw that a host should
have three servers in its list, too operate correctly. 

 

I have a system with a Trimble Acutime 2000 GPS attached to it (gps1). The
NTP server on this machine had three national references in its config, as
well as three local servers (time1,time2 and time3) with unknown
configuration.  Initially the local clients were configured to synchronize
to the gps1, time1, time2 and time3. What happened was that time1, time2 and
time3 decided to use gps1 as their time source. How does this affect the
accuracy of the local clients?  Based on this, the local clients were
instructed to only use gps1 as their time source. Is it terribly wrong to
configure the clients to only use gps1 as their time source? The clients are
one router and two switches away from the gps1, making it their closest time
source. And should the gps1 have other time sources than its GPS antenna? 

 

I did some measurements, where a host (running the NTP daemon) compared its
system time with the gps1's time every hour with 'ntpdate -d gps1'. And got
quite a strange behaviour (see table below), there were no errors reported
in the logs, which makes this behaviour even more strange. Apparently the
clock drifts, my question is why the there seems to be a (roughly) 6 hour
period, were the offset jumps a second? Both client and gps1 were running
Linux with a 2.4 kernel. 

 

Any Ideas? 

 

Best regards,

 

Patrik

 

 

Table column description: Sample is a running number, indirectly indicating
how many hours that has elapsed since start. Host time at the issuing of
ntpdate. Type and offset are reported from the ntpdate command. 

 

Sample:Host time:Type:Offset

1:1115278882:adjust:-0.00759

2:1115282482:adjust:-0.052178

3:1115286083:adjust:-0.104167

4:1115289683:step:0.952238

5:1115293283:step:0.835014

6:1115296883:step:0.759389

7:1115300483:step:0.699821

8:1115304083:step:0.646394

9:1115307683:step:0.596673

10:1115311283:step:1.6063

11:1115314884:step:1.51765

12:1115318484:step:1.4533

13:1115322084:step:1.39814

14:1115325684:step:1.34639

15:1115329284:step:2.38936

16:1115332884:step:2.28085

17:1115336484:step:2.20906

18:1115340084:step:2.1511

19:1115343685:step:2.09836

20:1115347285:step:3.13514

21:1115350885:step:3.05059

22:1115354485:step:2.96719

23:1115358085:step:2.90464

24:1115361685:step:2.85022

25:1115365285:step:2.79883

26:1115368885:step:4.23343

27:1115372485:step:3.70664

28:1115376086:step:3.65113

29:1115379686:step:3.59919

30:1115383286:step:3.54857

31:1115386886:step:4.6173

32:1115390486:step:4.4957

33:1115394086:step:4.41666

34:1115397686:step:4.35592

35:1115401286:step:4.30216

36:1115404887:step:4.25115

37:1115408487:step:5.27246

38:1115412087:step:5.17749

39:1115415687:step:5.11069

 

 

 

Patrik Arlos

Tech. Lic Telecommunications systems

PhD Candidate Telecommunications

Blekinge Institute of Technology

School of Engineering, Telecommunications Group

371 79  KARLSKRONA

SWEDEN

+46 (0)455 385654 Office

+46 (0)733 800312 Mobile

 




More information about the questions mailing list