Re: [chrony-users] Dealing with SNTP sources

[Harald Krammer <Harald.Krammer@xxxxxx>]

On Wed, 4 May 2011 13:40:12 -0700 (PDT), Bill Unruh <unruh@xxxxxxxxxxxxxx>
wrote:
> On Wed, 4 May 2011, Harald Krammer wrote:
> 
>> Am 2011-05-04 20:04, schrieb Bill Unruh:
>>> On Wed, 4 May 2011, Harald Krammer wrote:
>>>
>>>> Hello,
>>>> I have a question about dealing with SNTP sources.
>>>>
>>>> when I interact with two SNTP servers (also named as NTP server),
>>>
>>> An sntp server must, AFAIK be a stratum 0 server.
>>>
>>
>> Really? My opinion is, that the stratum behavior is equal. SNTP means 
a
>> usage of simple algorithm. e.g. only hard time set after a query
>> Some  SNTP implementation are extremely inaccurate.  In my area NTP is
>> also used over GPRS connections ( round trip delay till 100 ms - 2 s).
>> Accuracy does not play a major role in some applications.
> 
> That may be your opinion, but sntp has a specific definition within the
> standards and as far as I understand it, it is either a client only, or
it
> is
> stratum 0 server. It is not ever something which takes time from one
server
> and serves it to some other client.
> 

You are right. I also thought so, but in the field it's happen. My user
does 
not care about it, so there are my strange solutions. I implemented also  
IEC 60870-5-104 as time source for NTP. Our customers want such a thing.


> In fact, the staunch proponents (David Mills) would even regard chrony
as
> an
> sntp and not allow a machine running chrony to serve time to anyone
else. 
> (Of course what you do in the confines of your own little private net is
> up to
> you:-)

I did tested the different NTP implementations in my environment and 
chrony was my winner. I achieved an accuracy of + / - 500 us 
for 30 devices on a network with typical cpu and network load.  
Some customers want very precise times.

Target: Linux ARM 2.6.20 with uclibc / Cirrus EP9302 200 MHz 
and Linux ARM 2.6.20 with uclibc / PXA270 520MHz

I'm  very satisfied with Chrony. 
Thanks for the great work


> 
> Mills feels that he carefully designed ntp to give stable, reliable time
to
> others, and the simplifications stuck into sntp can easily destroy that
(
> as
> you have done by relabeling stratum) reliability/stability, and that
> allowing
> sntp into the time network would just bring it crashing down into chaos..
>   I have some sympathy for his position.
> 
> 

With the hacks in my area it's hard to find problems. Fortunately, 
they are not the standard for me - only exceptions.


Nice greetings,
Harald



---
To unsubscribe email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
with "unsubscribe" in the subject.
For help email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
with "help" in the subject.
Trouble?  Email listmaster@xxxxxxxxxxxxxxxxxxxx.