I deal with this problem by using a second, fast (LAN) network path and
sending periodic UDP "synchronization packets" between a timing server at
one end and a timing client at the other. The timing packets from the timing
server contain a 64-bit UTC timestamp. The timing client sets the time on
its machine to the UTC time from the packet. I send the packets about every
30 seconds. This keeps the clocks in synch and allows accurate one-way
transit time estimates. Without the packets the clocks drift enough to mess
up the measurement. I have tested wireless links (GPRS, EVDO) as well with
this method.
Gary Price
ICT
----- Original Message -----
From: "Yan Gu" <[EMAIL PROTECTED]>
To: <nistnet@antd.nist.gov>
Sent: Friday, September 29, 2006 6:26 AM
Subject: [nistnet] clock synchronization for NISTNet measurement
Hi,
I set up the Nistnet router with two machines connected to it.
A -->Nistnet router -->B
I want to measure one way delay from A to B and wonder how everyone is
dealing with the clock synchronzation problem on A and B.
I see that the one-way delay is less than 1 msecond between A and B so a
somewhat
accurate synchronization is preferred and I am not quite sure if NTP
itself is enough in this case.
Thanks
Yan
-Yan Gu ([EMAIL PROTECTED])
_______________________________________________
nistnet mailing list
nistnet@antd.nist.gov
http://www-x.antd.nist.gov/mailman/listinfo/nistnet
_______________________________________________
nistnet mailing list
nistnet@antd.nist.gov
http://www-x.antd.nist.gov/mailman/listinfo/nistnet
