[ntp:questions] An NTP Experiment

Matt Kinard kinard at raytheon.com
Wed Jul 20 19:54:42 UTC 2005


Folks,
   I've set up the following experiment:

I've got two Solaris 9 boxes running NTPv3.

I've got one box (hostname: aggie) set up as an NTP client to a pair of
local NTP servers via a LAN here on site (hence synchronization and
accuracy should not be an issue).

The other Solaris 9 host (named: raider) is set up as an NTP client to
aggie, but I've got it running through a FreeBSD host set up as a router
(hostname:  longhorn)

Effectivly, I've got the following configuration:

NTP Server -----------------FreeBSD Router-----------------NTP Client
  aggie
longhorn                                      raider

The caveat is i have the ablilty to introduce a variety of network
problems (delays/bandwith issues/etc...) on longhorn via the dummynet
utility.

I've written a script on longhorn that causes the network delay to move
around in a random fashion.

I can use xntpdc on raider to evalute the delay, offset, and dispersion.

Any suggestions as to what I could to to measure the delay's effect on
the overall accuracy on raider?

Thanks,
Matt



More information about the questions mailing list