[ntp:questions] Re: a couple of questions
Richard B. Gilbert
rgilbert88 at comcast.net
Thu Nov 4 16:32:11 UTC 2004
Folkert van Heusden wrote:
>I've got a couple of questions.
>1. what is the mtbf for a typical atomic clock? how often do these
>2. when one uses an atomic clock, should one still use 3
>of them because of fals tickers? or are atomic clocks reliable enough?
If you require the utmost reliability and accuracy, you should have at
least three. Most people, however, would consider it an extravagance!!
Cesium standards are EXPENSIVE!!!!!! If you buy three of the same make
and model and place them in service at the same time, it is likely that
they will all fail within a few months of each other!
>3. when using a pc as a kind of proxy, e.g.: it runs ntpd and connects
>to 4 upstream ntp-servers, and all clients connect to this one system.
>should one still preferably have 3 proxy-systems instead of 1? or only
>when the upstream link is likely to be broken in some cases?
If all proxy systems use the same upstream servers, the only benefit to
having more than one is the case where proxies fail.
>4. what is
>the mtbf of a typical gps receiver?
>5. if one has 3 time servers in an intranet, does it make sense to hook
>up a gps receiver to each of them, even if the gps receivers are
>installed at the same location on the roof of the building?
Two or more GPS receivers provide some protection against the failure of
one of them.
>6. what is the mtbf of a typical dcf77 receiver?
>7. if one has 3 time servers in an intranet, does it make sense to hook
>up a dcf77 receiver to each of them, even if the dcf77 receivers are
>installed at the same location in the building?
Two or more receivers protect against the failure of one.
>8. are there any documents describing best practices for setting up a time server for an
>intranet? i'm currently looking into an intranet with aprox 5000
>computers (windows/novell/linux/solaris) that need to sync themselves.
>8. about peering servers: should they peer if they use the same
>upstream-servers? or in any case?
I don't think there is any benefit.
You neglected to say what problem you are trying to solve (except in
VERY general terms). What degree of accuracy is required? What degree
of reliability is required?
I'd suggest at least three servers, each with a hardware reference
clock; perhaps one with a cesium or rubidium standard, one with a GPS
receiver, and one with a DCF77 receiver. If possible each server should
have serveral internet sources either as peers or as clients; at least
one source should be unique to each server. These servers should peer
with each other. These servers will all operate at stratum 1 and should
be dedicated to time service.
With five thousand clients, you may want to have some stratum 2 servers
dedicated to serving clients.
The design will depend strongly on your requirements and the available
More information about the questions