[ntp:questions] Geographical diversity of Stratum 1 servers

Richard B. Gilbert rgilbert88 at comcast.net
Tue Feb 19 20:45:25 UTC 2008


Peter Laws wrote:
> Peter Laws wrote:
> 
>> Looking for a reference on geographical diversity of Stratum 1 servers 
>> (GPS
>> appliances, in this case).  We have two clocks now (an old Datum TymServ
>> with a Symetricom sticker) and another unknown brand.  We'd like to add a
>> third at a completely different facility a mile or more away.
>>
>> What, other than the business continuity angle can I use?  I'm 
>> assuming it
>> would be better to have the clocks separated from a GPS perspective, too.
> 
> 
> Thanks to all that helped with this, here and off-line.  Turns out the 
> Datum TymServe 2100 (yep, that old) wasn't as dead as believed.
> 
> I've got it flashed to the last version of the fw and it's connected to 
> an antenna in another building and it is still, to my knowledge, working 
> (I forgot to get the port provisioned on the switch so I can't log in! 
> :-D ).
> 
> So, if it continues to function, it becomes the 3rd Stratum 1 server.
> 
> A follow-up if I may - it doesn't appear that I can peer the 2100s 
> together so my thinking is to have a group of Stratum 2s "around" the 
> three Stratum 1s.  They are all peered together and point to the 3 
> Stratum 1s for their time.  Everyone else then will point at the stratum 
> 2s.
> 
> Is there a better way to do this or will this work?
> 

Depending on the number of clients you need to support, you might not 
need a layer of stratum two servers.  If it's possible to configure this 
box to use broadcast mode, you could simply tell it to broadcast a "time 
hack" to your network every 64 seconds or whatever interval you find 
works best.  As each client initializes, it will query the server to get 
the round trip delay, crypto keys, etc.  Once the client has done its 
initial hand shaking, it will just sit there and listen. It's generally 
possible for an NTP server to support several hundred or even several 
thousand clients.




More information about the questions mailing list