[ntp:questions] NTP architecture recommendation
martin.burnicki at meinberg.de
Fri Aug 17 09:23:25 UTC 2007
> Well, I need to clarify "datacenter": actually these are not OUR
> datacenters, we use co-location at some provider, so we don't have a
> lot of control there (some of them will not even allow us to feed
> another cable into their server room). So even using a "remote
> antenna" (on the roof) might not be possible in some areas.
Do you know the Meinberg GPS antennas can be connected via a fiber-optic
Many data centers accept optical links to the outside world where they would
not accept copper cables.
Basically I agree to what other people already said: If possible, a local
reference clock is preferable to getting the time over a WAN link since you
never know about the quality and changing latencies of the WAN connection.
> It really
> looks like my approach will have to be: get a NTP-appliance with a
> reference clock (like the Meinberg devices) as close to the servers in
> the different sites as possible (ideally via a LAN).
If you install a LANTIME which only receives the time via NTP across a WAN
link then the performance may still be better than using a server machine
directly since the LANTIME includes a very good oscillator.
> Actually I will have to keep the servers in sync for 24 hours in the
> near future (for now 12 hours is ok; we are in financial business and
> will extend our operations to other markets soon, which basically will
> extend our "business hours" to 23 hours a day).
In fact you can just give it a try if you use 2 M300s with GPS receiver and
install it at different locations. On one M300 you can configure the other
device as a fallback time source, so both M300s have exact time since they
both sync to GPS, but one of them also polls the other one, so you can see
how good or bad the accuracy across the WAN link actually is.
More information about the questions