From owner-freebsd-questions Sat Jun 15 22:20:30 1996 Return-Path: owner-questions Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id WAA16642 for questions-outgoing; Sat, 15 Jun 1996 22:20:30 -0700 (PDT) Received: from pelican.altadena.net (pelican.altadena.com [206.16.90.21]) by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id WAA16630 for ; Sat, 15 Jun 1996 22:20:06 -0700 (PDT) Received: by pelican.altadena.net (Smail3.1.29.1 #10) id m0uVAFJ-0000RhC; Sat, 15 Jun 96 22:19 PDT Message-Id: Date: Sat, 15 Jun 96 22:19 PDT From: pete@pelican.altadena.net (Pete Carah) To: amengual@sadeya.cesca.es Subject: Re: Time Servers Newsgroups: freebsd.questions In-Reply-To: <31C19E19.1CA7@sadeya.cesca.es> References: Cc: questions@freebsd.org Sender: owner-questions@freebsd.org X-Loop: FreeBSD.org Precedence: bulk In article <31C19E19.1CA7@sadeya.cesca.es> you write: >Dave Babler wrote: >> How does one find available time servers for use with ntpdate? The only >> applicable address I've been able to find so far is the one given me by >> my ISP, but I'd like to use several to obtain a statistical average. Note that NTP doesn't do averaging for the actual time setting operation; it does some statistical filtering to choose which site to set from, but still only uses one at a time... It also assumes that your system clock runs at a pretty constant rate; that seems true for FreeBSD on the normal platforms but NOT for SGI Irix 5.2 (and seems as bad for 5.3, though they swore up & down that they would fix the time problems...) Some versions of sunos and solaris did time very badly on multiprocessor systems; 5.5 (2.5) seems pretty stable in most ways. Most of the freebsd systems I have anything to do with run 80-90 ppm slow. If you're using the kernel PLL it'll handle that fairly well, otherwise you may need tickadj. >You can find a list of secondary servers at: > >http://www.eecis.udel.edu/~mills/ntp/clock2.html >If you still cannot find a secondary server that is easily reachable from >your site, you could use a higher stratum server near your network (ask your >provider, for example). The secondarys in that list are spread all over the country; note that the net regions given for each site's connectivity are probably no longer current (UNLV is net-closer to us in Pasadena than UCLA, as are some bay area sites, for example; some texas sites are about the same). You could use one of ours (I try to keep them at st 3, though): ns.clubnet.net (MCI connection in Los Angeles; usually slow to cerfnet, ns3.clubnet.net sprint,and agis) ns.altadena.net (now MCI but probably soon AGIS) ns.interworld.net (AGIS connection in Los Angeles) ns2.interworld.net I don't block ntpq queries so you can find out what stratum we're at on each of those boxes at any given time. The system I'm writing from runs NTP at st 3 but shouldn't be used for clock setting as it's on a 28k connection with a (partial) news feed and has clock jitter of a few hundred millisecs with a 15-20 minute period (see our feed's nntpsend crontab entry :-). These connectivity problems may (hope, hope, hope) change if/when the politics surrounding mae-la ever get settled :-( One of our sites *may* acquire a GPS+PPS source pretty soon hung at the end of a T1 from a routing site. This will set the computer clock to higher precision than it can support :-) (see w3iwi's articles on the Motorola PPS card; it gives 40-50ns jitter in position-hold mode if enough birds are visible. For about half that price there is a Garmin card-level unit with a PPS output (but the serial is NMEA only; that is still adequate to set the clock to much higher precision than the computer+os can use.)). >This other URL points to useful NTP information: > >http://www.eecis.udel.edu/~ntp/ > >It seems to be the best NTP Web page available now. > >Hope this helps. It should... For more particular info on time in general, there are the NBS site and the USNO site (the latter is http://www.usno.navy.mil/ and follow links to the "directorate of time". I don't remember the NBS one but it shouldn't be hard to find.). -- Pete