[ntp:questions] Quality vs. Quantity

Danny Mayer mayer at ntp.org
Mon Mar 24 04:26:27 UTC 2014


On 3/23/2014 9:43 AM, steven Sommars wrote:
> Background:
> 
> NIST operates a DNS load balancer for NTP:  time.nist.gov
> See http://tf.nist.gov/tf-cgi/servers.cgi
> 
> USNO operates a server load balancer for NTP.  See for example:
> http://tycho.usno.navy.mil/ptti/2010papers/paper9.pdf
> 

That's a misconception. While I trust Richard Schmidt in what he says,
that's is not what you think he says. A DNS server can only respond with
a list of IP addresses and the normal design of most users is to take
the first one in the list. That's why most DNS servers will do
round-robin of the list, and is certainly true of BIND and Microsoft's
DNS servers. However an NTP server (and just about every application
that uses DNS) usually takes the first one and holds onto it for the
life of the application. In NTP we have started to take a different
approach and the pool option will use all of the returned IP addresses.
On the drawing boards is the idea that if a server doesn't respond after
a while the address can be dropped and another DNS query is done to get
a new set of addresses to be used.

On the NTP inference engine side, keeping the same address allows it to
stabilize since if you get different answers from what is claimed to be
the same address you will be receiving entirely diffeent timestamps that
will have that address with wildly fluctuating information and that will
always get dropped as a candidate for a truechimer.

Danny

> 
> 
> On Sun, Mar 23, 2014 at 2:43 AM, Terje Mathisen <terje.mathisen at tmsw.no>wrote:
> 
>> Daniel Quick wrote:
>>
>>> While this should be obvious, I always have to ask how and why...
>>> While considering that the number of requests to our time servers
>>> will grow over time since the client decides which server to sync
>>> with.
>>>
>>> Do we want a Netspeed setting that assists with taking the load off
>>> some of the more heavily, higher-speed servers? or do we want to keep
>>> a setting where we serve fewer clients with the highest resolution of
>>> time given specific setup and let the client queries grow from there?
>>> I suppose this also takes into the smart dns load-balancing that goes
>>> on in the background.
>>>
>>
>> You really do NOT want load-balancing of ntp servers!!!
>>
>> Put them all in a pool and let the clients connect to all, distributing
>> the load automatically.
>>
>> Terje
>>
>>
>>> Any input would be appreciated.
>>>
>>> Thanks,
>>>
>>> Daniel
>>>
>>>
>>
>> --
>> - <Terje.Mathisen at tmsw.no>
>> "almost all programming can be viewed as an exercise in caching"
>>
>>
>> _______________________________________________
>> questions mailing list
>> questions at lists.ntp.org
>> http://lists.ntp.org/listinfo/questions
>>
> _______________________________________________
> questions mailing list
> questions at lists.ntp.org
> http://lists.ntp.org/listinfo/questions
> 
> 



More information about the questions mailing list