Re: [Linuxptp-devel] [PATCH 0/4] Leap second fix and improvements.
On 9/15/2021 3:08 AM, Miroslav Lichvar wrote: > I was improving the coverage of the linuxptp testsuite and noticed a bug > that needs to be fixed before the next leap second. That's the first > patch. The others are improvements to better see what is happening > around leap second, make the grandmaster flags reliable, and better > handle leap seconds on a PHC synchronized to UTC. > The whole series makes sense to me. Reviewed-by: Jacob Keller Thanks, Jake > Miroslav Lichvar (4): > clock: Accept new UTC offset after leap second. > clock: Print info message when leap flags change. > clock: Clear leap flags after leap second. > clock: Notify servo about leap second on UTC hardware clock. > > clock.c | 75 + > 1 file changed, 60 insertions(+), 15 deletions(-) > ___ Linuxptp-devel mailing list Linuxptp-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxptp-devel
Re: [Linuxptp-devel] [PATCH] Adding Percentile Filter Support
On 9/11/2021 1:51 AM, joseph.matan...@gmail.com wrote: > From: Joseph Matan > > The percentile filter can be very useful when running in a non-ptp-aware > environment. > For example, if we set a low percentile value (and the filter length is large > enough), > we can still get a good estimation of the real delay, > even if our setup is running under heavy network traffic. Ok. > (when setting the percentile value to 0.50, the median filter is a private > use-case of the percentile filter) > What does this part mean? ___ Linuxptp-devel mailing list Linuxptp-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxptp-devel
Re: [Linuxptp-devel] [PATCH] Adding Percentile Filter Support
On Tue, Sep 21, 2021 at 12:31:57PM +0300, Joseph Matan wrote: > > > > Enabling the weighted filtered mode can help, but not as much as > > disabling the filter completely, at least from what I have seen in my > > tests. > > > I guess it depends on the environment (devices and switches in the between > and traffic), > but from my experience I've received better results when I worked with the > percentile filter in weighted filtered mode, Can you show your results? Here is my test comparing the median filter with 0.1 percentile and no filter in a network with an asymmetrically distributed delay (mean 10 microseconds). https://fedorapeople.org/~mlichvar/tmp/ptp/delayfilter.png As you can see, the percentile filter added a significant asymmetry of about 4 microseconds. In terms of RMS time error, it's about 3x worse than the median filter and about 5x worse than no filter. The shared part of the config was: delay_filter_length 11 pi_proportional_const 0.04 pi_integral_const 0.00024 > and even much better results when I used my own weighted filter > implementation (which I also hope to share soon). Well, in that case I think these changes would better be submitted and reviewed together. -- Miroslav Lichvar ___ Linuxptp-devel mailing list Linuxptp-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxptp-devel
Re: [Linuxptp-devel] [PATCH] Adding Percentile Filter Support
> > Enabling the weighted filtered mode can help, but not as much as > disabling the filter completely, at least from what I have seen in my > tests. > I guess it depends on the environment (devices and switches in the between and traffic), but from my experience I've received better results when I worked with the percentile filter in weighted filtered mode, and even much better results when I used my own weighted filter implementation (which I also hope to share soon). > With the percentile filter you expect an asymmetric distribution. The > reported delay is more stable, but the clock is less accurate. What is > the use case here? > Make the clock accuracy more robust in a non-ptp-aware environment (this will also require the use of weighted filtered mode). > Why not just disable the filter? > As I said before, because I've received better results when I worked with the percentile filter in weighted filtered mode (and much better results with my own weighted filter implementation). In general, the percentile filter will keep the old behaviour of the median filter, but will add additional flexibility for users who may need it. If we are ok with this, I will fix the patch according to your comments and resend it. Thanks, Joseph ___ Linuxptp-devel mailing list Linuxptp-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxptp-devel