On Tue, Feb 18, 2014 at 07:08:41PM +0100, Delio Brignoli wrote:
>
> I also assumed the peer delay had to be non-negative to be meaningful.
> Would you be OK with a patch that introduced a minimum acceptable peer
> delay configuration option?
I think even for gPTP, the measured mean peer propagation can
legitimately be less than zero. Let me explain. First, consider the
performance limits from the 802.1AS-2011 standard.
** B.1.1 Frequency accuracy
The fractional frequency offset of the LocalClock relative to the
TAI frequency (see Annex C) shall be within ±100 ppm.
** B.2.3 Pdelay turnaround time
The pdelay turnaround time is the duration of the interval
between the receipt of a Pdelay_Req message by a port of a
time-aware system, and the sending of the corresponding
Pdelay_Resp message.
The pdelay turnaround time of a time-aware system, measured
relative to the TAI frequency (see 8.2), shall be less than or
equal to 10 ms.
Let us assume that the "responding" node has a 10 ms turn around time,
but that its clock is fast by 100 ppm. It will report the interval
(t3-t2) in Figure 11-1 as having a duration of 10.001 ms. If the
the "requesting" node clock is syntonized to TAI, and the propagation
delay is 100 nanoseconds, then that node will report (t4-t1) with a
duration of 10.0002 ms. The result then is a propagation time of -400
nanoseconds.
I think we have to allow this, or maybe rethink our peer delay
measurement for gPTP. In any case this patch is not correct, since it
will prevent the port from being labeled as asCapable, and it also
prevents discovering the neighborRateRatio.
It is perfectly fine for the measured propagation to be negative as
long as the calculation for the master offsets are correct after
applying the neighborRateRatio.
Thanks,
Richard
------------------------------------------------------------------------------
Managing the Performance of Cloud-Based Applications
Take advantage of what the Cloud has to offer - Avoid Common Pitfalls.
Read the Whitepaper.
http://pubads.g.doubleclick.net/gampad/clk?id=121054471&iu=/4140/ostg.clktrk
_______________________________________________
Linuxptp-devel mailing list
Linuxptp-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linuxptp-devel
