Regarding NTP configuration ntp.conf
arpit...@emerson.com
I wish to know that If I am providing 2 server names in ntp.conf
without "prefer" option then with which server my system will sync the
time.
I googled it I came to know that the server which has less "jitter" it
will sync with. Other thing is both the servers are at same strand
value.
Can anyone please explain it. I will be thankful to you .
Thanks
Arpit
Kevin Oberman
> From: <arpit...@Emerson.com>
> Sender: questions-bounces+oberman=es....@lists.ntp.org
I think you probably meant stratum instead of "strand". In any case, the
algorithm is a bit more complex than "jitter", but it is the dominant
contributor. I'm not entirely sure what you are asking, though. It sound
like you found the answer.
I will say that having two servers is probably the worst case. You
really want three, five, or seven. Those allow for "good" servers to
out-vote a bad server. If you only have two, there is no indication as
to what the "best" time is. The jitter may be low, but the time may be
off by a lot. If you have three servers, if one goes bad, ntp will ignore
it and always pick from the servers that have about the sime time.
--
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: obe...@es.net Phone: +1 510 486-8634
Key fingerprint:059B 2DDF 031C 9BA3 14A4 EADA 927D EBB3 987B 3751
Richard B. Gilbert
>> Date: Thu, 17 Dec 2009 14:24:56 +0800
>> From: <arpit...@Emerson.com>
>> Sender: questions-bounces+oberman=es....@lists.ntp.org
>>
>> Hi,
>> I wish to know that If I am providing 2 server names in ntp.conf
>> without "prefer" option then with which server my system will sync the
>> time.
>> I googled it I came to know that the server which has less "jitter" it
>> will sync with. Other thing is both the servers are at same strand
>> value.
>> Can anyone please explain it. I will be thankful to you .
>
> I think you probably meant stratum instead of "strand". In any case, the
> algorithm is a bit more complex than "jitter", but it is the dominant
> contributor. I'm not entirely sure what you are asking, though. It sound
> like you found the answer.
>
> I will say that having two servers is probably the worst case. You
> really want three, five, or seven. Those allow for "good" servers to
> out-vote a bad server. If you only have two, there is no indication as
> to what the "best" time is. The jitter may be low, but the time may be
> off by a lot. If you have three servers, if one goes bad, ntp will ignore
> it and always pick from the servers that have about the sime time.
Three is not a good number of servers. If any one of the three fails
either by not responding or by serving incorrect time you are reduced to
two servers; back to the worst case. The magic numbers are four, five,
or seven, protecting you from the failure of one, two, or three,
respectively, of the configured servers.
Brian Utterback
> Three is not a good number of servers. If any one of the three fails
> either by not responding or by serving incorrect time you are reduced to
> two servers; back to the worst case. The magic numbers are four, five,
> or seven, protecting you from the failure of one, two, or three,
> respectively, of the configured servers.
Even without a failure, three servers do not always allow two servers
to out vote a bad third one. Usually yes, but not always, depending on
the different dispersions and offsets. Four will always allow three to
out vote a bad fourth.
Evandro Menezes
>
> I will say that having two servers is probably the worst case. You
> really want three, five, or seven.
Actually, one wants "3x + 1" (Byzantine fault-tolerance) servers in
order to out vote "x" bad servers, i.e., 4, 7, 10, etc.