Impact of ARP lookup on local NTP traffic

[Gary E. Miller]

Yo Dan!

On Mon, 30 May 2016 14:53:31 -0500
Dan Drown <dan-ntp at drown.org> wrote:

> https://dan.drown.org/bbb/arp-impact/ tries to show the impact of
> ARP on the request latency.  Response isn't impacted because the
> peer learns the ARP entry on the request.

Very interesting data.  Almost all the time it makes no difference, but
everyone in a while it removes spikes up to 600 uSec!

I have changed some "server" statements to "minpoll 4".  That changes
the NTP polling interval from 64 seconds to 32 seconds.  Since an
ARP entry is elligible for eviction at 60 seconds that is a nice
cushion.

After an hour I do not notice much change in jitter, but I do see
improvement in offset.  I think the time constant for the offset is
longer than for the jitter, so it holds on to one-off errors longer.

I have added a section on this to the time service howto.

Thanks for brining this up.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
	gem at rellim.com  Tel:+1 541 382 8588
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 473 bytes
Desc: OpenPGP digital signature
URL: <http://lists.ntpsec.org/pipermail/devel/attachments/20160530/3c763c3e/attachment.bin>