[newbie]: How to determine accuracy
ihs...@my-deja.com
I'm setting up a ntp server in my lan (stratum two
on a sunos 4.x machine) and would like to find
how to determine the accuracy of my time.
Is there a specific command which I can use ?
any input is really appreciated.
tq,
-ihsan
Sent via Deja.com http://www.deja.com/
Before you buy.
Ulrich Windl
> Hi,
> I'm setting up a ntp server in my lan (stratum two
> on a sunos 4.x machine) and would like to find
> how to determine the accuracy of my time.
> Is there a specific command which I can use ?
[x]ntpdc -c sysinfo
Look for "stability" ("frequency") and "jitter". Smaller values are
preferrable. If you have kernel support, look for "estimated error"
with "ntpdc -c kerninfo"
Ulrich
Frederick Bruckman
ihs...@my-deja.com writes:
> I'm setting up a ntp server in my lan (stratum two
> on a sunos 4.x machine) and would like to find
> how to determine the accuracy of my time.
> Is there a specific command which I can use ?
ntpq -c rl -c pe
The second part especially, the ntpq "peers" command, will tell
you how close you are keeping to your stratum one servers.
ihs...@my-deja.com
thanks for the explaination. Shall I say the
smaller the offset is better ?
But how about the offset which is sometimes
below 0 (negative).
What does it means ?
tq,
--ihsan
In article <7u2cj0$18np$3...@news.enteract.com>,
Frederick Bruckman
ihs...@my-deja.com writes:
> But how about the offset which is sometimes
> below 0 (negative).
> What does it means ?
It means that you are running _fast_, that is,
that server is behind you by that amount.
Hans Lambermont
In article <7u2cj0$18np$3...@news.enteract.com>, f...@enteract.com (Frederick
Bruckman) writes:
> In article <7u1og0$f7s$1...@nnrp1.deja.com>,ihs...@my-deja.com writes:
>> I'm setting up a ntp server in my lan (stratum two on a sunos 4.x
>> machine) and would like to find how to determine the accuracy of my
>> time. Is there a specific command which I can use ?
>
> ntpq -c rl -c pe
>
> The second part especially, the ntpq "peers" command, will tell you
> how close you are keeping to your stratum one servers.
But this does not tell you *how* accurate you are, as it is a *measure*
for your accuracy. I would like to know the maximum deviation in ms from
UTC to within 95%. Can this be done ?
(See also my previous posting about this subject
<93965241...@dibbs1.mpn.cp.philips.com>)
regards,
Hans Lambermont
--
ir. Hans Lambermont, Architect/Developer, Hans.La...@nl.origin-it.com
Origin IT MS/INS, Intranet Services tel/fax: (+31 40 27) 85376/88729
PO Box 218, 5600MD Eindhoven, Netherlands VN507 gsm: (+31) 6 53252172
David Schwartz
Hans Lambermont wrote:
> But this does not tell you *how* accurate you are, as it is a *measure*
> for your accuracy. I would like to know the maximum deviation in ms from
> UTC to within 95%. Can this be done ?
Yes, but you need a local source that's closer to UTC than the
deviation you are trying to measure. If you don't have one, you are
asking for the impossible. How can you tell how accurate a clock is
without access to a more accurate clock?
DS
Hans Lambermont
In article <38076A4D...@webmaster.com>, David Schwartz
<dav...@webmaster.com> writes:
> Hans Lambermont wrote:
>> But this does not tell you *how* accurate you are, as it is a *measure*
>> for your accuracy. I would like to know the maximum deviation in ms from
>> UTC to within 95%. Can this be done ?
>
> Yes, but you need a local source that's closer to UTC than the
> deviation you are trying to measure.
No. (But it would be nice to have of course, you can get a more precise
maximum deviation from UTC.)
> If you don't have one, you are asking for the impossible. How can you
> tell how accurate a clock is without access to a more accurate clock?
By adding the known maximum errors in the chain. But i find this
mathematics rather 'nasty' to set up (from what i remember from a course
years back at the university :-). But i am very sure it can be done
nicely.
Our setup is a unix-based ntp server getting its time from 4 GPS
receivers, each placed in a continent, with their own ip/ntp stack. From
the latter the maximum deviation is known (one claims 250 nanoseconds of
UTC for hardware and software combined, for example). Then ntpd on the
unix machine should be able to estimate its maximum accuracy error
relative to the GPS receivers. But i still don't know how to get this
data from ntp's logfiles.
David Schwartz
> By adding the known maximum errors in the chain. But i find this
> mathematics rather 'nasty' to set up (from what i remember from a course
> years back at the university :-). But i am very sure it can be done
> nicely.
>
> Our setup is a unix-based ntp server getting its time from 4 GPS
> receivers, each placed in a continent, with their own ip/ntp stack. From
> the latter the maximum deviation is known (one claims 250 nanoseconds of
> UTC for hardware and software combined, for example). Then ntpd on the
> unix machine should be able to estimate its maximum accuracy error
> relative to the GPS receivers. But i still don't know how to get this
> data from ntp's logfiles.
>
> regards,
> Hans Lambermont
This is not what the original poster wanted. Yes, it's easy to get
measurements of the type "we believe it is no worse than this". What is
wanted, however, is "95% of the time, it's known to be as good as this."
DS