[ntp:questions] ntp architecture

konsu konrad.azzopardi at gmail.com
Mon Oct 25 14:45:42 UTC 2010


On Oct 25, 4:13 pm, unruh <un... at wormhole.physics.ubc.ca> wrote:
> On 2010-10-25, konsu <konrad.azzopa... at gmail.com> wrote:
>
> > Dear ntp group,
>
> > I am quite new to ntp and am designing an ntp architecture in a 3
> > datacentre environment but there are so many possibilities that I
> > would like to get an opinion.
>
> > We have three datacentres {all within the same country} and in one of
> > the datacentres there is a GPS antenna to which I intend to connect an
> > appliance that provides ntp service.
>
> That appliance would presumably be a computer, not running windows but
> linux or bsd.
>
> That GPS should have a PPS, since the nmea sentences are notoriously bad
> for timekeeping ( unless you consider 10s of milliseconds OK).
>
>
>
> > A server in each of the three datacentres, would act as stratum 2
> > clocks peering each other and synchronizing to my ntp GPS appliance
> > {which can be called stratum 1, I guess}.
>
> You should have a backup or three-- possibly outside your organization,
> for use when the gardener snips your GPS lead in wire, or some kids
> steal the antenna.
>
>
>
> > Are there any guidelines about best practices and if an architecture
> > like this would look good ? Then the intention is that all Linux/UNIX
> > servers would sync time to the stratum 2 clocks {3 in total}. If
> > possible can someone give me examples of config files for each of the
> > 3 stratum 2 servers ?
>
> Why not just have all reference the GPS source? How many clients are
> there?
>
>
>
> > Many thanks in advance
> > Konrad

Hi, thanks for the reply..........................

Indeed the appliance is a Meinberg LANTIME M300/GPS which has an
inbuilt GPS clock and provides an ntp daemon { I am sure it is BSD or
linux inside}. since we will have over 400 clients I preferred to use
a layered approach and doing a stratum 2 layer seems easy and elegant.
As you say I am now thinking what happens if the appliances fails , I
can either have a procedure to switch to ntp pool project {only as a
disaster scenario until replacement} or else I have the option that
our datacentre supplier offered a DCF77 signal to which I can also buy
another appliance so effectively I would have two stratum 0 and
stratum 1 sources but as far as I understand 2 servers as sources
should never be used so I don't know if if I should stick to one
source and if it fails I failover to the other system {radio or ntp
pool project only for disaster}

tnx
Konrad




More information about the questions mailing list