[ntp:questions] Time sync on a windows domain
David J Taylor
david-taylor at blueyonder.not-this-bit.nor-this-part.co.uk
Thu May 24 14:32:49 UTC 2007
clarkec8 at hotmail.com wrote:
> I'm investigating different methods of keeping a network in time. We
> currently have a network with 6000 XP clients, 200 2003 servers and
> about 20 Solaris/Unix servers.
> We currently use ntpdate on the unix servers and w32time on the
> windows server with the PDC going to the internet for time via NTP.
> Our current time difference accross the network is about 2 seconds.
> This seems a lot to me. We've also go some 2003 servers that swing
> around all over (around 4 mins). These are pre SP1 though and I
> believe there were some fixes in SP1.
> I would like to get a GPS NTP device and sync to that and I'd like
> some advice.
> 1. If I use the PDC to point to the NTP device and use the domain
> heirarchy what sort of accuracy can I expect?
> 2. If I still use w32time but set it to NTP and point all clients at
> the clock device what accuracy could I expect?
> 3. Should I use something else entirely and ignore w32time?
> I know I'm asking on an NTP forum and if I asked on a windows one I
> may get different advice but I'm trying to get a spread of opinion.
> I've searched everywhere for accuracy on w32time but Microsoft don't
> seem too keen on giving out such knowledge.
My answer would be: "Why not run NTP everywhere?", i.e. your option 3.
Having said that, I have heard that the latest W32time (for Windows 2003
only) is not as bad as previous versions in being closer to NTP
performance, but I have no direct experience of this.
If you only run NTP, then there is only one software to support, albeit in
both UNIX and Windows variants.....
My own, much simpler network, performs as show here. One FreeBSD server
synced to a GPS 18 LVC source, and a mixture of Windows PCs. About
100msec or better accuracy.
More information about the questions