On 01/02/15 22:37, jdow wrote:
Nahhh. Unless 'chrony' has completely mucked up the NTP spec, it deals
with symmetrical propagation delays pretty well. It effectively
reocords when you sent the request, when the response arrived, and
what the server thought the time was to derive the relevant skews.
On many satellite ISP links I've heard that they are extremely
asymmetrical with regards to the link delay. In these cases downlink
to the customer is satellite and uplink is one form or other of
dialup. I've no idea how frequent they are these days. I heard of them
more than a decade ago in reference to the Hughes "stuff". NTP becomes
quite inaccurate in those cases.
{^_^} Joanne
The Viasat system uses radio up and down, the subscriber has a ~2 watt
transmitter, doesn't sound like much but it works well.
For what it's worth I just ran five tests at http://speedof.me/ and saw
the following "latency" measurements:
841, 1652, 870, 1683, 874 ms.
I don't know how well chrony or ntp can deal with those but I suspect it
may do pretty well if we are considering a few ms of error acceptable? I
believe the delay is mainly transit time up and down at both ends of the
circuit and should be nearly constant.
I also have the Viasat voip service and that has no noticeable delay
during normal conversation, prefer it to my cell phone, certainly voice
quality is better and I have even used it to send medical data from a
heart monitor, "set the device on the phone and press send." In the past
attempts at using Skype, etc. were poor due to confusion caused by the
delay.
I will have to look for a gps device of some sort for comparison ...
Bob
--
http://www.qrz.com/db/W2BOD
box10 Fedora-21/64bit Linux/XFCE
