[ntp:questions] Re: asyncronous path between "time client" and ntp server

Danny Mayer mayer at gis.net
Thu Oct 13 00:49:58 UTC 2005


Bob Beers wrote:
> On 10/10/05, Danny Mayer <mayer at gis.net> wrote:
> 
>>Bob Beers wrote:
>>
>>>BTW, I really meant asymmetrical, not asynchronous, but the former
>>>would likely imply the latter.
>>>
>>
>>Not at all. They're unrelated and orthogonal to each other. How do you
>>know that the path is asymmetric? Do you have any evidence to back this
>>up? If so can you estimate the amount of this asymmetry? There is at
>>least one handle you can use to tell ntp about the asymmetry but I'd
>>have to look it up. It's more important to look at the evidence first.
>>
>>Thanks again for the replies. The more questions I ask the more
> 
> questions I need to ask. I may have to quit while I'm behind. :)
> 
> I used traceroute from ntp server to ntp client and vice versa to
> come to the conclusion that the path between the two was not symetrical.
> Do you need to see the traces? There are several identical 192.168.20.
> subnets being NAT'd, and a 192.168.0. subnet where the ntp server
> is being NAT'd, but this subnet is not behind a satellite link.
> 
> Anyway, there has been an alternate time server provided which is not
> NAT'd. (I hope it is ntpd, but I haven't confirmed)
> 

Yes, I seem to recall that satellite links are always asymmetric with 
outbound request going via land and the incoming return coming in via 
satellite. If you can figure out the difference in the delay you can add 
a fudge to the server to account for this. I'll let someone else tell 
you exactly what you need to do.

Danny

PS Please tell gmail to turn off interpretation of DNS addresses. Plain 
text is preferred in a mailing list like this. Not everything is a URL.



More information about the questions mailing list