[ntp:questions] Re: NTP client on Windows platform provides less accurate results then on the UNIX or Linux. Why?
mayer at ntp.isc.org
Sun Apr 2 18:18:54 UTC 2006
> On 4/1/06, Danny Mayer <mayer at ntp.isc.org> wrote:
>> Why? There is very rarely a need to reboot a Windows machine. 99% of all
>> hotfixes should not require a reboot. I probably reboot once every 3
>> months. Only badly written installers require reboots.
> I agree, but tell that to Microsoft and other Windows software vendors.
> Many of the more recent MS application patches don't requrie reboots,
> but it seems at least one of the patches every 2nd Tuesday of the month
> does despite the fact that they don't patch the kernel, HAL, or any
> other "critical" subsystem. All IE-related patches still require a
> restart. And MS isn't the only offender; I recently had to bounce all
> of my Windows servers for a Symantec patch.
Personally, apart from work-related requirements, I never touch IE. That
said I still get caught out by the MS patch-of-the-minute syndrome.
> If you're not applying the MS patches, well, that's your choice I
> suppose. If it's a server running very few applications, and you're
> sure no junior network admin will ever browse the web from it, there's
> proabably no real reason for the majority of MS patches. But most
> people who deal with network security believe in defense-in-depth,
> which means not just relying on a perimeter firewall and permissions to
> secure a machine.
You wouldn't believe some of the stuff I run on my machines! However,
where I can I avoid all MS products. More than anything else it's
because MS products are such big targets for virus writers.
>> NTP will always perform worse in comparison with a
>>> Linux box that is always on.
>> Most people running NTP on Windows would disagree with you.
> You're misunderstanding me. All I meant was that a machine that is
> rebooted frequently will never run NTP as well as well as an NTP box
> that is not restarted frequently, regardless of the platform.
First of all one has nothing to do with the other. The biggest area that
Windows had problems was the interrupts due to running media apps. We
fixed this in the Meinberg release. The only reason it's not yet in the
regular distribution is because I've been chasing a high-priority
WIndows bug and I haven't had time to get it in.
>>> If your hardware has any power-management features...
>> That's almost unavoidable and is not just Windows.
> CPU clock-throttling features are typically not enabled by default on
> any OS *except* Windows and MacOS. Which is not a bad thing, as those
> power-saving features are a definite bonus for just about every
> application other than NTP.
>> All systems have firewalls, routers, switches, etc. There's no real
>> difference here.
> Yes there is. Very few non-Windows boxes run AV/IPS software that
> actively pattern-scans network connections as many of the popular
> client security packages do. Many organizations do this scanning at the
> gateway/firewall as well, which would affect all machines running any
> OS. But the client-machine AV scanning of network connections (not just
> file activity) is probably unique to Windows boxes (such as the client
> Windows workstations the original poster described).
That should have no more affect on NTP than anything else running on the
box. We have seen no such affects. If you want to report a problem then
you need to send much more detail.
> That said, I do have stable stratum-2 Windows servers running NTP that
> see ~1ms average offsets after running for a few days. This is about
> the same performance as my other stratum-2 NTP platforms. But overall,
> the Windows machines get restarted more frequently, and therefore are
> generally worse-performing as NTP servers.
Restarting the machine has no long-term effect. I am constantly
restarting ntpd as I spend a lot of time debugging and the main effects
I see is when I disconnect it from the network, but that's not surprising.
> Please note that I am most certainly not an anti-MSFT zealot. I manage
> a shop with 20+ servers, and all but three run Windows 2003 SP1. It's
> the best choice for a huge number of applications. And ntpd *can* run
> servicably on it; I use the Meinberg distribution on many machines
> myself. But I believe the original poster was talking about poor ntpd
> performance on a bunch of Windows workstations. Who knows what sort of
> activity those workstations see. At the very least, they need to reboot
> for all of the IE-related patches. And aside from patches, those
> workstations are probably restarted more frequently that servers,
> whether they need it or not.
I'm really not sure why you think that restarts have such a great effect
on NTP. It's designed to converge quickly and certainly no more than
about a day.
More information about the questions