Re: [chrony-users] Comparing timekeeping performance with chronyd to ntpd

[Miroslav Lichvar <mlichvar@xxxxxxxxxx>]

On Tue, Apr 30, 2019 at 10:49:29AM -0700, Watson Ladd wrote:
> On Tue, Apr 30, 2019 at 5:15 AM MUZZULINI Frank
> <Frank.MUZZULINI@xxxxxxxxxxxxxx> wrote:
> >
> > Well, no daemon can be more precise than its sources.
> > I would use independent measurements against the same source(s) individually using "ntpdate -q <addr>" and/or "chronyd -Q -f <addr.cfg>" (*). Preferably both.
> 
> That's our plan, but we're not sure what sort of analysis to do on the
> results. (We will be using a GPS slaved rasbperry pi to gather the
> results).

Are you interested in measuring the stability/accuracy of the system
clock on the machine running ntpd/chronyd, or stability/accuracy of
the time it provides as an NTP server? That would be two different
things.

A RPI synchronized to PPS for testing an NTP server or client should
be accurate to about 20 microseconds. If you need more accurate
measurements, you would need something with hardware timestamping and
ideally a PPS input on the NIC itself (e.g. I210/I211) rather than a
serial port.

Here are some measurements I did with a RPI and other hardware:

https://fedorapeople.org/~mlichvar/ntpserver/

-- 
Miroslav Lichvar

-- 
To unsubscribe email chrony-users-request@xxxxxxxxxxxxxxxxxxxx 
with "unsubscribe" in the subject.
For help email chrony-users-request@xxxxxxxxxxxxxxxxxxxx 
with "help" in the subject.
Trouble?  Email listmaster@xxxxxxxxxxxxxxxxxxxx.