Previous part

From: Casper.Dik@Holland.Sun.Com (Casper H.S. Dik - Network Security Engineer) [-/+]
Date: 17 Apr 1998 08:04:53 GMT
[-/+]
Newsgroups: comp.protocols.time.ntp,comp.unix.solaris,alt.solaris.x86
Subject: Re: Solaris 2.6 kernel PLL and xntpd
[-/+]
X-Keywords: bug
[-/+] PLL [-/+]

[[ PLEASE DON'T SEND ME EMAIL COPIES OF POSTINGS ]]

Dima Volodin <dvv@dvv.ru> writes:

>Has anybody made it work? Has anybody found any use for Solaris's kernel
>PLL at all?

It seems to only work with Sun's own xntpd?

There's a bug:  4095849

Casper
--
Expressed in this posting are my opinions.  They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.


From: Dima Volodin <dvv@dvv.ru> [-/+]
Date: Sat, 18 Apr 1998 13:20:44 -0400
[-/+]
Newsgroups: comp.protocols.time.ntp,comp.unix.solaris,alt.solaris.x86
Subject: Re: Solaris 2.6 kernel PLL and xntpd
[-/+]
X-Keywords: PLL
[-/+]

Per Hedeland wrote:

> In article <casper.892800205@uk-usenet.uk.sun.com>
> Casper.Dik@Holland.Sun.Com (Casper H.S. Dik - Network Security Engineer)
> writes:
> >Dima Volodin <dvv@dvv.ru> writes:
> >
> >>Has anybody made it work? Has anybody found any use for Solaris's kernel
> >>PLL at all?
> >
> >It seems to only work with Sun's own xntpd?
>
> Um, no - Sun's xntpd works OK-ish, but it doesn't *use* the kernel PLL -
> don't know if that's because it isn't supported in that old version
> (3.4y) or just not (auto-)configured in.

The funny thing is that there is a reference to ntp_adjtime in Sun xntpd's
dynamic symbols table, but no matter how I tried I've never been able to make
it actually call ntp_adjtime.

> --Per Hedeland

Dima


From: per@erix.ericsson.se (Per Hedeland) [-/+]
Date: 18 Apr 1998 10:48:54 GMT
[-/+]
Newsgroups: comp.protocols.time.ntp,comp.unix.solaris,alt.solaris.x86
Subject: Re: Solaris 2.6 kernel PLL and xntpd
[-/+]
X-Keywords: bug
[-/+] dispersion [-/+] panic [-/+] PLL [-/+]

In article <casper.892800205@uk-usenet.uk.sun.com>
Casper.Dik@Holland.Sun.Com (Casper H.S. Dik - Network Security Engineer)
writes:
>Dima Volodin <dvv@dvv.ru> writes:
>
>>Has anybody made it work? Has anybody found any use for Solaris's kernel
>>PLL at all?
>
>It seems to only work with Sun's own xntpd?

Um, no - Sun's xntpd works OK-ish, but it doesn't *use* the kernel PLL -
don't know if that's because it isn't supported in that old version
(3.4y) or just not (auto-)configured in.

>There's a bug:  4095849

That's the one you told me about in private mail, which might cause a
panic - I haven't heard of anyone experiencing that though, the problem
we see when using kernel PLL with recent versions of xntpd is that
things never stabilize - high offset and dispersion values, stepping,
etc.

--Per Hedeland
per@erix.ericsson.se


From: dalton@cup.hp.deletethis.com (David Dalton) [-/+]
Date: 17 Apr 1998 17:05:19 GMT
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: Adjusting time on mission critical systems
[-/+]
X-Keywords: SLEWALWAYS
[-/+]

Daniel Wong (wongd@syr.lmco.com) wrote:

:>Does any one know if there is a simpler way to turn the initial jump off
:>and force slewing of the clock?

The compile option -DSLEWALWAYS does this.  It effectively removes the 128
millisecond limit, so you will never make a step.  It might take you a long
time to converge initially, however.  The max slew rate varies from one CPU
to the next (due to the crystal), but my workstation will only slew about
40 microseconds per second.  If I were initially off by 5 minutes, my
workstation would take about 125 minutes (over two hours) just to get close,
and probably twice that long to stabilize.  So there are tradeoffs....
TANSTAAFL

HP provides this SLEWALWAYS behavior in PHNE_12689, but it is not for
everyone.  A few customers (like the original poster here) are extremely
allergic to any step in time.  Most people would do better with the
standard NTP in PHNE_11019.

--
-> My $.02 only   Not an official statement from HP {They make me say that}
--
    As far as we know, our computer has never had an undetected error.
---------------------------------------------------------------------------
David Dalton           dalton@cup.hp.deletethis.com            408/447-3016


From: colin@nyx10.nyx.net (Colin Plumb)
Date: 17 Apr 1998 23:56:39 GMT
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: Adjusting time on mission critical systems
[-/+]
X-Keywords: adjustment
[-/+] configuration [-/+] fudge [-/+] HP-UX [-/+]

In article <6h0c12$jp4$1@ins8.netins.net>,
Rodney J. Jenkins <jenkins@netins.net> wrote:
> I currently have about 50 HP-UX machines.  These machines are using the NTP
> protocol to keep their time syncronized.  The NTP time server is currently
> using its internal clock for a time source.
>
> I have aquired 3 GPS time clocks to act as time soruces.  I understand how
> to configure the three servers and point all of the clients to those
> servers.
>
> The question:
> What is the best way to adjust the time from the bad time to GPS time on the
> current server?
>
> The constraints:
> I need to adjust the time back 00:02:56. (HH:MM:SS)
> All servers need to have the same time.  (Accurate to the same second)
> The applications cannot have the time adjusted backwards by any amount.
> I cannot shut the applicaitons down.

There is a way to tell NTP to slew always, so it will not step the time.
Given your constraints, I'd have this compiled in always.

To keep everything in sync during the changeover, install the GPS
clock in *one* place (probably your existing master) and let it slew,
dragging everyone else with it.

This will take some time; indeed it might be best to let the master
take even more time than necessary to make sure that none of the clients
fall off the steep slope.

For example, you could use the line

fudge 127.127.t.u time1 176

which would add 176 seconds (check with someone else about the sign
convention; you might want -176!) to the time you get from the clock.
Then ratchet it down by one second per hour or some such.)

Oh, BTW, to *test* your GPS configuration, just set the GPS receiver to
a higher stratum than your current local clock so it will be trusted
less.  You'll still get figures from it that you can examine, but
they'll be ignored.

Once you have everyone within a second or so of UTC, then set up the other
two GPS receivers on hosts on different parts of the network and set up
your network in a nice fault-tolerant way.

But to get it started, it's easiest to make it non-fault-tolerant so that
you can have a nicely controlled fault at a given location.

If you use the adjust-the-fudge technique, then it is possible to use
multiple GPS receivers from the start and just restart them all
(so they'll re-read the config file) at the close to the same time when
you make an adjustment.  But it seems more complicated than necessary.
--
        -Colin


From: moshier@mediaone.net () [-/+]
Date: 18 Apr 1998 19:07:31 GMT
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: local ATOM?
X-Keywords: adjustment
[-/+] configuration [-/+] delay [-/+] resolution [-/+] rubidium [-/+]

On Wed, 25 Feb 1998, Bruce Bartram 303-497-6217 wrote:

> I think that you can look at the "xntpdc -p" and showp <server> info
> to see if the network delay is slowly increasing over the period.  I
> don't think the internet/NTP is good to a few ms, but I doubt it has
> any long term drift.
>
> Perhaps the clock divider sequence in the PC doesn't make the exact
> required divider ratio and the ticks are slightly off the nominal
> size.  I think that linux allows some method of tweaking the host
> clock rate by changing the exact amount of time the OS thinks
> happens between ticks and this could be the "freq pot".  After such
> OS adjustment, the last little bit could be inserted in ntp.drift
> to get to the ppb'ish range, if the OS tweak doesn't have the needed
> resolution.

Using a calculated value in ntp.drift for the rubidium PC, the local
clock has drifted about a millisecond after six weeks.  I think this
is a nice testimonial for linux; it does not seem to be dropping clock
interrupts and there was only one minor scale factor bug/feature
encountered while figuring the clock rate.

The table summarizes the four internet clocks monitored by the
free-running local clock.  Comparing what I get with the cable TV
modem and some other megabit T1 internet connections, it's apparently
very lucky to find one with a sub-millisecond scatter.

This experiment was terminated by a power failure courtesy of the
local electric company.  After rebooting the computer I included the
26.8 microseconds per day in ntp.conf, so the rate now ought to be
within about one second per thousand years.  It's very unlikely that
the computer will be up for as long as even one year at a stretch, so
the correct strategy for this hardware configuration is now clear: the
uncertainty of internet clock readings is so high that there is no
point in synchronizing the local atomic clock to the erroneous
external net clock readings.  The local clock can be set to within a
few milliseconds from the net, and the free-running drift will be less
than that before the next time the clock has to be set again.

Clock    Regression Line   Post-fit   Number of
        usec, usec/day    RMS Error  Data samples
                          usec
-------  ---------------   ---------  ------------
harvard    803 + 26.8 d        828      1728
columbia  -261 + 36.9 d       2814       550
nearnet  -1617 - 94.8 d       4935       386
mediaone  1680 - 62.0 d       4622      1048


From: Philippe Kipfer <philippe-kipfer@hp.com>
Date: Tue, 21 Apr 1998 13:55:39 +0200
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: The use of SNTP
X-Keywords: RFC
[-/+] SNTP [-/+]

Micha/l Lipi/nski wrote:

> Hi,
>
> I need some information on SNTP:
>
> Where is the Simple Network Time Protocol (SNTP) used to synchronize time?
> What sort of applications need synchronization and SNTP could be suitable
> for them?
> Where can I find information on computer time synchronization with SNTP?

The SNTP RFC's references were already given in an other answer.
A typical use of SNTP is PC synchronization from an Unix environment :-)
I made good experiments with the freeware called Dimension 4 (before joining
HP :-).
This software is available under
http://www.thinkman.com/~thinkman/dimension4/index.htm
Other clients software references can be found at
http://www.eecis.udel.edu/~ntp

If you use "dimension 4" you will probably configure it in a way that the
PC take its time
at boot time or for example every hour. At the synchronization time, the
local clock is
brutally updated by a time step. This is certainly acceptable at boot time,
but maybe not
if you run a DB application :-)

My 2cent worth, hope this help.

--
Best regards, Ph. Kipfer (SSSO-Bern)

------------------------------------------------------------------------
  Hewlett-Packard (Switzerland) AG     Phone:    +41 (0)31 980 31 11
  Philippe Kipfer                      Fax:      +41 (0)31 980 33 90
  Meriedweg 11                         Telnet:   780-3352
  CH-3172 Niederwangen/Bern            Internet: philippe_kipfer@hp.com
  X400:S=Kipfer/G=Philippe/C=CH/A=400NET/P=HP/O=HP/
      OU1=HP-Switzerland/OU2=OM3
========================================================================


From: wiu09524@rrzc4 (Ulrich Windl) [-/+]
Date: 23 Apr 1998 11:52:12 GMT
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: ntp clock updates
X-Keywords: maxpoll
[-/+] minpoll [-/+] poll [-/+]

In article <353E3998.52B2990@cisco.com> derrot <sshen@cisco.com> writes:

> new to ntp.
> was wondering if anybody knew what the algorithm for
>  calculating frequency of clock updates?

Yes, it's also knows as the polling rate. You can bound it within
minpoll and maxpoll. Usually starting at minpoll, the polling interval
will increase if the clock is stable reative to its synchronization
source.

in xntpdc's loopinfo command it's the number after "poll adjust":
Positive is good, negative is bad. +30 will increase the polling
interval (I think).

>
> does the client request updates? currently, i have a very simple
> config- just a client getting time from stratum-2 clock.

Have a look at http://www.eecis.udel.edu/~ntp for more information to
get you started.

Ulrich


From: wiu09524@rrzc4 (Ulrich Windl) [-/+]
Date: 23 Apr 1998 11:54:55 GMT
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: What is different XNTP and NTP?
X-Keywords: implementation
[-/+]

In article <6hmu08$8qd$1@news.kornet.nm.kr> "CoalAsh" <coalash@voin.com> writes:

> Please, Answer...
>
>

Maybe next time supply some more information in the mesage body...

NTP = Netwaork Time Protocol
XNTP = popular implementation of the above

Ulrich


From: teltrust.com <teltrust@xmission.xmission.com>
Date: 23 Apr 1998 19:32:55 -0600
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: ntpdate while ntpd is running?
[-/+]

Chris Johnson <cjohnson@palomine.net> wrote:
: Yes. ntpdate won't mess with your clock if xntpd is in charge. For debugging
: you can use ntpdate -d, which will tell you what ntpdate would have done,
: but won't actually make any changes to your clock, and there won't be any
: complaints about the NTP socket being in use.

Thanks.

Yet another question: I'm getting timeouts with ntptrace when I hit a
publicly available time server:

# ntptrace ntp.someuniv.edu
XYZ.SOMEUNIV.EDU: stratum 3, offset 39.782415, synch distance 0.01563
dns3.someuniv.edu:  *Timeout*

Is this because its stratum 2 server isn't public, and its ignoring me?

Also, with a ntp server in my /etc/ntp.conf file, if I do a ntptrace on
localhost, I just get a timeout error (but it takes several seconds to
timeout). With just my local clock, it comes back just fine.

-Trevor


From: "Chris Johnson" <cjohnson@palomine.net>
Date: Thu, 23 Apr 1998 18:14:27 -0400
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: ntpdate while ntpd is running?
[-/+]

>I was wondering if its normal for ntpdate to complain that the ntp socket
>is in use when ntp is already running:
>
>23 Apr 15:50:07 ntpdate[22690]: the NTP socket is in use, exiting

Yes. ntpdate won't mess with your clock if xntpd is in charge. For debugging
you can use ntpdate -d, which will tell you what ntpdate would have done,
but won't actually make any changes to your clock, and there won't be any
complaints about the NTP socket being in use.

Chris Johnson


From: Stefan Monnier <monnier+comp/os/linux/development/system/news/@tequila.cs.yale.edu>
Date: 23 Apr 1998 15:05:32 -0400
[-/+]
Newsgroups: comp.os.linux.development.system,comp.protocols.time.ntp
Subject: Re: ntp and Linux Pentium SMP
X-Keywords: adjtimex
[-/+] PLL [-/+]

Patrick Cipiere <Patrick.Cipiere@sophia.inria.fr> writes:
> We also have a bunch (~10) of SMP Pentium (same hardware, + 1 processor)
> On these machines the clock is not stable.

Have you tried to look at the output of `adjtimex -print' ??
On my machines (with hand compiled xntpd of various versions as well as with
the contrib xntpd-3.5.91 rpm module), this clearly showed that xntpd wasn't
actually using linux's PLL code.  Ever since I switched to Debian's xntpd
package, the time has been kept much better.
Now `much better' doesn't mean perfect.  I'm pretty sure that there is
something wrong with SMP and system clock (whether it's a Linux problem
or not I don't know) but at least with a good xntpd it seems that the
problems have mostly gone away.

        Stefan


From: Reynir Siik <reynir@royal.net> [-/+]
Date: Thu, 23 Apr 1998 21:41:13 +0200
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: xntp and Year 2000
X-Keywords: FAQ
[-/+]

Terrance Ingoldsby wrote:
>
> Does anyone know if there is a statement indicating whether xntp (and
> the NTP protocol) is Y2K compliant?  I can't see any reason why it
> wouldn't be, but our organization (and probably many others), is trying
> to get an "official" statement on every s/w product that will be in use
> on or after Dec 31, 1999.
>
> I checked the FAQ, but didn't see anything there.
>
> --
>
> --------------------------------
>  Terrance Ingoldsby
>  (403)541-4824
>  Terrance.Ingoldsby@pipe.nova.ca

Hi,

Check out this site: http://www.eecis.udel.edu/~mills/y2k.htm
It might solve your problem.
NTP will have it's overflow sometime in the year 2036. So ther will be
plenty of time to take action.

Cpt Reynir Siik
Swedish Armed Forces


From: Harlan Stenn <stenn@whimsy.udel.edu> [-/+]
Date: 24 Apr 1998 17:30:24 -0400
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: ntp-4.0.72f is in /testing/
X-Keywords: bug
[-/+] DES [-/+] Mills [-/+] PLL [-/+]

4.0.72f is available on ftp://ftp.udel.edu/pub/ntp/testing/ .

This version is suitable for export.

DES support is now optional, disabled by default, and there is no DES
code in the distribution.  See README.des for details on how to get DES
support working, if you *need* it.

The ntp_*time kenel calls are not used under Solaris2.6 at the moment.

As I understand it:

- ntp_gettime works just fine
- ntp_adjtime is Broken in the kernel
- There is probably a bug in the kernel regarding the PLL code

- Apparently, both Dave Mills and some Sun kernel dudes are currently
  actively working on a fix, and there is light at the end of the
  tunnel.

Dvae has made a substantial number of (undocumented) changes since 72d.
72e was very short-lived.

Sorry, no diffs are available.  I'm hoping to cut 4.0.73 in a week or
so, assuming all we get between now and then is a trickle of reports.

There will be a diff from 4.0.72 to 4.0.73.

If we can find a source-code-control package that's so easy to use that
Dave will use it, we'll make the (read-only) source tree available to
folks.  We're currently evaluating Perforce (www.perforce.com) and CVS
(perhaps with cvsup).

H


From: Jim Dennis <jadestar@rahul.net>
Date: 25 Apr 1998 12:30:01 GMT
[-/+]
Newsgroups: comp.unix.admin,comp.protocols.time.ntp
Subject: Time Sync for Disconnected LAN?
X-Keywords: FAQ
[-/+] poll [-/+] UDP [-/+]

All,

I've just spend a couple of hours browsing the man pages for ntp,
xntp, timed, adjtime, and a few FAQ's and web pages to boot.

All I'd like is to keep my systems with a second or so (per week
or month) of the stratum 2 time servers nearest me).  Time
"chimeheads" (as a few of the participants in the archives I
read identify themselves) seem to bring obfuscation to a new art.

I'm on a small disconnected LAN (ethernet).  All are currently
Linux boxes (I have one Mac that's not even turned on and some
of these are dual boot --- but none have seen their alter egos
(MS OS personas?) in a long time).

One of the systems, betelgeuse, is the router between my private
LAN segment an my ISDN router.  That system provides proxies
(Squid, DeleGate) and sometimes IP masquerading for the rest
of the LAN (which uses a 192.168.x.* RFC1918).  It seems natural
enough to make this a local time server ('timed -M'?) and have it
periodically poll a couple of stratum 2 servers (I have four from
California listed).

It would be nice if my polling didn't actually bring up a link
but only used the link when it was already up.  I'm sure that's
possible in diald and with ppp links (since I could just configure
it to poll at least once per day when the link is going up).
However, with my ISDN link it all happens transparently to
betelgeuse.

I gather from one note that I might consider just periodically calling
'/sbin/clock -s'  or  '/sbin/clock -a' to set the system time from the
CMOS clock (I guess the built-in hardware is more accurate than the
system software).  I guess I could then configure betel to check
once a week or so.

If I understand it correctly I could then just run timed -M on
betel, and just timed on the other local systems.

In any event it would seem that we have an embarrassing number of
choices when it comes to time/clock synchronizations of Unix
systems.  I see /sbin/clock, NTP, xntp, netdate, timed, adjtime
(which I gather is supposed to be called by the other subsystems)
and at least four different entries in the /etc/{services,inetd.conf}
(not counting the TCP/UDP duplicates).


From: kees@echelon.nl (Kees Hendrikse) [-/+]
Date: Sat, 25 Apr 1998 12:45:46 GMT
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: ntp stuff on www.echelon.nl
X-Keywords: loopstats
[-/+] SCO [-/+]

I put up a few ntp-related pages on our web site, including the recipes
for making xntp run on SCO Unix, binaries for SCO Unix (for those without
the development system) and an couple of perl scripts for loopstats
statistics. See: http://www.echelon.nl/en/ntp/

--
Kees Hendrikse                               | email:     kees@echelon.nl
                                            | web:        www.echelon.nl
ECHELON consultancy and software development | phone: +31 (0)53 48 36 585
PO Box 545, 7500AM Enschede, The Netherlands | fax:   +31 (0)53 43 37 415


From: rbthomas@lilypad.rutgers.edu (Rick Thomas) [-/+]
Date: 28 Apr 1998 17:05:58 -0400
[-/+]
Newsgroups: comp.protocols.time.ntp,comp.os.vxworks
Subject: Re: Opposing time syncs.
X-Keywords: dialup
[-/+] reset [-/+]

David Uhrenholdt <duhrenho@vette.sanders.lmco.com> writes:

>...occasionally I get a pair of
>time resets of the form

>time reset  (step) -0.977381 s
>followed about a minute and a half later with
>time reset  (step) 0.978809 s

>What might be causing this?

You are probably seeing occasional short-lived large asymmetric delays
(and/or other kinds of flakeyness) in the internet connection between
your machine and the time-server.  Configuring more time-servers might
help - but if the problem is in a link (such as a dialup PPP link)
that all client<=>server paths share in common, it won't help at all.
Path-diversity is the answer, if you can get it.

Rick


From: poli@removeme.poli.net
Date: Wed, 29 Apr 1998 17:08:43 -0600
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: sntp support in Cisco IOS 11.2?
[-/+]
X-Keywords: Cisco
[-/+] configuration [-/+] SNTP [-/+]

In article <3546E4E9.E8AA4B40@whu-koblenz.de>#1/1,
  Nils-Henner Krueger <nhk@whu-koblenz.de> wrote:
>
> The cisco guys (comp.dcom.sys.cisco) don't know a solution, maybe
> you can help me?
>
> Does anybody know what happened to sntp support in Cisco IOS 11.2?
>
> I'm running a Cisco 1003, and the "Configuration Fundamentals
> Configuration Guide" says (p. 312):
> "Simple Network Time Protocol (SNTP) is a simplified, client-only
> version of NTP for use on Cisco 1003 ... routers."
>
> But when I go to configuration mode and enter "sntp server ip-no", I
> simply get "% Unrecognized command".
> Anybody with more success?
>
> nhk

Nils,

      I'm running an IOS 11.2(5)P, and NTP is just working fine.

      To get it working, first login as enable, then enter the config mode,
and then, type "ntp ?" and you'll get help in NTP.

Regards,

        //\/\\arco //>oli

-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/   Now offering spam-free web-based newsreading


From: kees@echelon.nl (Kees Hendrikse) [-/+]
Date: Wed, 29 Apr 1998 12:38:04 GMT
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: sntp support in Cisco IOS 11.2?
[-/+]
X-Keywords: configuration
[-/+]

In <3546E4E9.E8AA4B40@whu-koblenz.de> Nils-Henner Krueger writes:

> But when I go to configuration mode and enter "sntp server ip-no", I
> simply get "% Unrecognized command".

As far as I know none of the ISDN-only type cisco routers ever had NTP
support in any form.

--
Kees Hendrikse                               | email:     kees@echelon.nl
                                            | web:        www.echelon.nl
ECHELON consultancy and software development | phone: +31 (0)53 48 36 585
PO Box 545, 7500AM Enschede, The Netherlands | fax:   +31 (0)53 43 37 415


From: Reynir Siik <reynir@royal.net> [-/+]
Date: Wed, 29 Apr 1998 18:50:23 +0200
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: sntp support in Cisco IOS 11.2?
[-/+]
X-Keywords: Cisco
[-/+] configuration [-/+] SNTP [-/+]

Nils-Henner Krueger wrote:
>
> The cisco guys (comp.dcom.sys.cisco) don't know a solution, maybe
> you can help me?
>
> Does anybody know what happened to sntp support in Cisco IOS 11.2?
>
> I'm running a Cisco 1003, and the "Configuration Fundamentals
> Configuration Guide" says (p. 312):
> "Simple Network Time Protocol (SNTP) is a simplified, client-only
> version of NTP for use on Cisco 1003 ... routers."
>
> But when I go to configuration mode and enter "sntp server ip-no", I
> simply get "% Unrecognized command".
>
> Anybody with more success?
>
> nhk

Hi,
I have configured several Cisco 3620 routers with IOS 11.2(10p) and the
command for this is:

ntp server xxx.xxx.xxx.xxx

This makes the router work in client only mode. I'm not familiar with
the Cisco 1003 router, but it may be worth trying. Or maybe you already
have.

Good Luck

Cpt Reynir Siik
Swedish Armed Forces
Staff and Communications School


From: kees@echelon.nl (Kees Hendrikse) [-/+]
Date: Wed, 29 Apr 1998 20:27:55 GMT
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: ntpdate for SCO unix
X-Keywords: SCO
[-/+]

In <35474246.FB0CE3FB@tro.net> Hein Bloed <ata@tro.net> writes:

> thx... but i need a precompiled binary, because on our SCO Server there
> is _no_ compiler installed... :(
>
> u know where to get it?!?

Same place. http://www.echelon.nl/download/ntp/sco5xntp.tar.gz has
the binaries, including ntpdate.

--
Kees Hendrikse                               | email:     kees@echelon.nl
                                            | web:        www.echelon.nl
ECHELON consultancy and software development | phone: +31 (0)53 48 36 585
PO Box 545, 7500AM Enschede, The Netherlands | fax:   +31 (0)53 43 37 415


From: cbbrowne@news.brownes.org (Christopher B. Browne)
Date: 30 Apr 1998 04:10:36 GMT
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: Re: ntp or xntp, how accurate?
X-Keywords: precision
[-/+] synchronized [-/+]

On Wed, 29 Apr 1998 14:03:24 +0200, Hans Berglund <hb@spacetec.no> posted:
>I have not (yet) used ntp, and I wonder how accurate two computers
>can be synchronized.
>
>Do we talk about seconds, milliseconds or microseconds?
>
>Is ntp the best solution, or does it exist more accurate way to do
>time synchronization?

Precision will vary based on:
a) How accurate the clocks available on the two computers are,
b) The speed of the network connection between the two computers, and
c) The speed of the connection between the computers and whatever outside
time source is being used.

If the clocks on the computers are not very accurate, then accuracy will be
pretty limited, regardless of anything else.

If they are "decently connected" to the each other and to external time
sources, then accuracy will be pretty good.

I have three hosts on my Ethernet LAN that I can keep synchronized to within
milliseconds of one another within the LAN; I'm afraid I don't get great
synchronization between them and the rest of the world, as the clock on the
host that has the Internet connection has a fairly large "deviance." It
moves off by probably a minute or so per day, which pushes all the local
clocks off.

When I've got a connection "live" to the Internet, I keep to within a few
hundredth's of a second of "true" time.  When the connection shuts off for a
few hours, precision slides by a few seconds.

How good your results will be will depend on all of these sorts of things.

If you have a continuous Internet connection at 56K or better that *both*
hosts can get at you should be able to get *accuracy* of on the order of
milliseconds.

If the hosts are on an Ethernet segment, then you should get a couple of
additional orders of magnitude of *precision*; those hosts should be able to
stay within 0.0001s of one another, if they have half-decent clocks on
them...

--
Those who do not understand Unix are condemned to reinvent it, poorly.
-- Henry Spencer          <http://www.hex.net/~cbbrowne/lsf.html>
cbbrowne@hex.net - "What have you contributed to Linux today?..."


From: Terje Mathisen <Terje.Mathisen@hda.hydro.com> [-/+]
Date: Thu, 30 Apr 1998 20:38:06 +0200
[-/+]
Newsgroups: comp.protocols.time.ntp
Subject: [++]Re: Netware time sync to a NTP server??
[-/+]
X-Keywords: adjustment
[-/+] firewall [-/+] implementation [-/+] SNTP [-/+]

Gary L. Zeiger wrote:
>
> Does anyone know of a Netware .NLM that will sync the server time to a
> NTP server that is running on my NT firewall?  I have been using
> RDATE.NLM,
> but that will only sync to a Unix server.  Thanks.

Andre Schilds (from Switzerland) have written SNTPCLNT.NLM, a client
only implementation of the SNTP protocol.

        http://www.neatech.ch/sntpclnt/

It is not free, but registration was very cheap, something like $15-20.

I use this as the master reference for a 300+ server NetWare 4.11
network.

The current version does not try to adjust the tick rate, so it tends to
do quite a few adjustments to the server clock, but he is currently
testing a version that does adapt itself to the average of the sources.

I am using this on a test server, after about 24 hours it had adjusted
the server clock to the point where it is running very nicely:

Since 09:56 this morning (about 10.5 hours) it has drifted
(free-running) about 15-20ms, corresponding to about 40ms/day, whereas
it used to drift 17+ seconds/day before SNTPCLNT started the adjustment
process.

Terje

PS. I have not verified that his SNTP implementation works with the NT
version of NTP, but I have no reason to believe otherwise.

--
- <Terje.Mathisen@hda.hydro.com>
Using self-discipline, see http://www.eiffel.com/discipline
"almost all programming can be viewed as an exercise in caching"


Main part