Re: [Ntp] WGLC: draft-ietf-ntp-interleaved-modes

[Miroslav Lichvar <[e-mail]>]

On Fri, Dec 14, 2018 at 01:03:29PM -0800, Harlan Stenn wrote:
> - 18ns corresponds to a precision of about -26.  While that's pretty
> special,

I'm not sure if that is really special. Modern CPUs have a time-stamp
counter, which can be read in few tens of nanoseconds. The precision
of a typical hardware clock on NIC is same or even better.

> I remain curious about how long the delay is between pulling
> the transmit stamp and the packet going out over the wire.

It should be similar to receiving a packet, i.e. few microseconds.
Depending on the OS/driver, with a very fast NTP server close to the
client, it is actually possible to receive the response before the
transmit timestamp of the request.

> If the basic
> precision of the system clock is only -19, we're not going to do better
> than 2 microseconds.  For situations like this, I'm really curious if
> we'll even see any benefit to client/server interleave mode.

I think an improvement in accuracy should be possible even with such a
clock.

> - I'd like to explore a "follow up" message as an alternative.

A protocol using follow up messages could certainly be specified, but
it would enable traffic amplification. That's probably the last thing
we want to add to the client/server mode. There are few sentences
about this in the latest version of the draft.

In any case, such a protocol would be very different from the
interleaved modes and should be specified separately. It could work in
all three modes: client/server, symmetric, broadcast.

-- 
Miroslav Lichvar