[ntp:questions] Re: isolated network + accuracy

Richard B. Gilbert rgilbert88 at comcast.net
Mon Dec 13 18:47:21 UTC 2004


Alaios wrote:

>Thx really for the answer... If i have understand
>correctly the problem of not having good accuracy is a
>problem of my server because it always loses
>interrupts.
>
>If i connect the server to the internet under a
>stratum 2 server, do u think that the server can
>achieve accuracy below 1 ms? Do u think i must
>configure the server in a specific way for my
>requirements?
>
>After finishing with the server then the clients can
>achieve synchronization with accuracy of 1 ms or this
>is still impossible? The major problem is that i want
>all the clients to be synchronized and each of them
>has an accuracy of 1 ms or less.
>Really thank u!
>Have a nice day 
>
>--- "Richard B. Gilbert" <rgilbert88 at comcast.net>
>wrote:
>
>  
>
>>Alaios wrote:
>>
>>    
>>
>>>Goof morning. I have an isolated network and i want
>>>      
>>>
>>to
>>    
>>
>>>achieve accuracy smaller than 1ms. Do u know if
>>>      
>>>
>>that
>>    
>>
>>>is possible? How i must configure the clients for
>>>that?
>>>Any suggestion?
>>>
>>>
>>>		
>>>__________________________________ 
>>>Do you Yahoo!? 
>>>Yahoo! Mail - Find what you need with new enhanced
>>>      
>>>
>>search.
>>    
>>
>>>http://info.mail.yahoo.com/mail_250
>>> 
>>>
>>>      
>>>
>>Without a stable source of time, I'd say that
>>synchronization within 1ms 
>>is not possible.  The typical, unsynchronized,
>>computer clock is neither 
>>stable nor accurate.
>>
>>You will need a hardware reference clock of some
>>sort.   A cesium 
>>frequency standard (with clock option) is the best
>>there is.  It's also 
>>extremely expensive.  A rubidium standard is second
>>best and 
>>considerably cheaper but still expensive to buy and
>>maintain.  A GPS 
>>timing receiver can be obtained for less than $500
>>US and is a very good 
>>choice if you can place an antenna where it has an
>>unobstructed view of 
>>the entire sky.  An oven controlled quartz crystal
>>oscillator can 
>>provide a stable source.
>>
>>Configure one computer with attached hardware
>>reference clock as the 
>>server.   Configure the clients to get their time
>>from the server.
>>
>>Don't expect much from Microsoft Windows in the way
>>of accuracy; the 
>>clock resolution is too low and the operating system
>>tends to lose 
>>interrupts (clock ticks) when it is busy doing
>>something else.  Linux 
>>also has been known to have problems with losing
>>interrupts.
>>_______________________________________________
>>questions mailing list
>>questions at lists.ntp.isc.org
>>https://lists.ntp.isc.org/mailman/listinfo/questions
>>
>>    
>>
>
>
>
>		
>__________________________________ 
>Do you Yahoo!? 
>Yahoo! Mail - Helps protect you from nasty viruses. 
>http://promotions.yahoo.com/new_mail
>  
>
If your server loses interrupts, I'd say you have no hope of achieving 
accuracy!

Using an internet server as a time source will almost certainly not 
allow you to achieve 1ms accuracy!  10ms accuracy is possible over the 
internet but cannot be guaranteed.

Many many computer operating systems update their clocks at intervals of 
10ms or longer.   In a system updating at 10ms intervals, the time 
reported at any random instant can be wrong by as much as 9.9999999999. 
. . ms and, on average, will be within 5ms of the correct time.

1ms accuracy is a very high standard and quite difficult to meet.  I 
don't believe that  Windows is capable of it.  Linux may not be 
either.   Some flavors of Unix are capable, at least in principle.  I 
believe that Solaris is.  I lack sufficient recent experience with other 
flavors of Unix to offer an informed opinion.



More information about the questions mailing list