[ntpwg] Documents, slides, etc. from WG meeting

Thu Oct 25 09:26:53 UTC 2007

Haiko and all,
I am not making distinction between SW-based and HW-based routers. 

My point is that IPv4 routers must comply with RFC 1812 including, among other things, the "local delivery" decisions as defined in section 5.2.3. Unicast PTP packets that must be forwarded by a compliant router clearly fall into the "queue for forwarding to the appropriate next hop" from the point of view of these requirements. The only modifications of the packet in this case would be decrement of IP TTL and the subsequent IP header checksum re-calculation.

Beyond that, the only standard-defined way to assure that transit routers look at the contents of a packet beyond its IP header (not to mention modifying anything in the packet) that I am aware of is the Router Alert option as defined in RFC 2113. A well-known example of a protocol that requires such look-up and modification is RSVP (RFC 2205). And it explicitly states that its "Path, PathTear, and ResvConf messages must be sent with the Router Alert option" (section 3.3).

Hence IMHO, for Unicast PTP to run over an IP network and, at the same time, to benefit from TC capabilities, the Router Alert option must be included into the IPv4 header of the PTP packets - and that in addition to any actual HW or FW that would really perform the TC function.

Of course, if there is no TC clocks in the IP network to be traversed by unicast PTP, the Router Alert option can be (and probably, should be) omitted. 

My 2c.
                    Sasha

________________________________

From: Heiko Gerstung [mailto:heiko.gerstung at meinberg.de]
Sent: Wed 24-Oct-07 18:56
To: Sasha Vainshtein
Cc: stbryant at cisco.com; mayer at ntp.isc.org; NTP Working Group; Brad Knowles; Israel Sasson; Alon Shtern; Osnat Shasha
Subject: Re: [ntpwg] Documents, slides, etc. from WG meeting

Sasha,
I am sorry but my reference to "special hardware" in the context of TC
should have been "special firmware" and was based on the assumption that
routers are hardware devices. It might be enough to modify ntpd to work
in a "forwarding mode" on such a router - instead of the current BC-like
operation (get sync from one side and sync others on the other side).

Best Regards,
 Heiko

Sasha Vainshtein schrieb:
> Heiko,
> Thank you for a prompt response.
> However, the question was not so much about special HW as about the
> need to modify the protocol to make the routers on the way to inspect
> the PTP packets they normally should not look at beyond their IP headers.
> 
> Regards,
>                     Sasha
> 
>
> ------------------------------------------------------------------------
> *From:* Heiko Gerstung [mailto:heiko.gerstung at meinberg.de]
> *Sent:* Wed 24-Oct-07 02:57
> *To:* Sasha Vainshtein
> *Cc:* stbryant at cisco.com; mayer at ntp.isc.org; NTP Working Group; Brad
> Knowles; Israel Sasson; Alon Shtern; Osnat Shasha
> *Subject:* Re: [ntpwg] Documents, slides, etc. from WG meeting
>
>
> Sasha Vainshtein schrieb:
> > Heiko, Stewart and all,
> > I just wonder if the PTP "transparent clock" (TC) mechanism is not in
> > contradiction with the way IP routers are supposed to work?
> > When PTP runs over IP, at least some of the PTP traffic (DELAY_REQ and
> > DELAY_RESP messages ) are unicast UDP/IP Packets running between the PTP
> > Slave and PTP Master with the well-known UDP port assigned to PTP. Hence
> > the "normal" routers would simply pass these packets thru without even
> > looking at the UDP ports not to mention UDP payload (i.e.., PTP proper).
> > Of course one may add the "router alert" option to these packets (even
> > if, AFAIK, PTP does not specify that), but even this would hardly help
> > since the router would probably simply put the packet on the "slow
> > path".
>
> That is correct, a TC concept always would include special hardware which
> supports this mode. Just like hardware timestamping would require
> special NICs.
>
>    Heiko
>
>
> >
> > My 2c
> >                  Sasha
> >
> >
> >
> > -----Original Message-----
> > From: ntpwg-bounces+sasha=axerra.com at lists.ntp.org
> > [mailto:ntpwg-bounces+sasha=axerra.com at lists.ntp.org] On Behalf Of Heiko
> > Gerstung
> > Sent: Wednesday, October 24, 2007 4:06 PM
> > To: stbryant at cisco.com
> > Cc: mayer at ntp.isc.org; NTP Working Group; Brad Knowles
> > Subject: Re: [ntpwg] Documents, slides, etc. from WG meeting
> >
> > Stewart Bryant schrieb:
> >> Heiko
> >>> The one thing that really enables PTP to reach an impressive accuracy
> >
> >>> level over an Ethernet/UDP connection is the fact that it uses
> >>> hardware time stamping and therefore requires specialized hardware
> >>> (if you want to benefit from that superior level of accurcacy).
> >>> 
> >> There is nothing to stop NTP doing h/w timestamping per se. The
> >> critical difference is that PTP (1588)  does on path correction
> >> through the use of the TC mechanism. If TC were added to NTP the two
> >> protocols would provide the same level of accuracy.
> >
> > I agree. The TC concept introduces quite a number of security problems,
> > which have been addressed in V2 but obviously lead to an increased
> > complexity. Adding a few things like hw-timestamping and on path
> > correction to NTP would result in a synchronization protocol which
> > offers a very good performance over both a WAN link and a controlled LAN
> > environment (plus compatibility to millions of existing clients). Well,
> > I have to go and talk to my engineers now :-)
> >
> > Best Regards,
> >   Heiko
> >
> >
> >
> >> Stewart
> >>
> >
> >
> > --
> > ------------------------------------------------------------------------
> >
> > *MEINBERG Funkuhren GmbH & Co. KG*
> > Lange Wand 9
> > D-31812 Bad Pyrmont, Germany
> > Tel.: ++49 (0)5281 9309-25
> > Fax: ++49 (0)5281 9309-30
> > eMail: heiko.gerstung at meinberg.de <mailto:heiko.gerstung at meinberg.de>
> > Internet: www.meinberg.de <http://www.meinberg.de/>
> >
> > ------------------------------------------------------------------------
> >
> > Meinberg radio clocks: 25 years of accurate time worldwide
> >
> > _______________________________________________
> > ntpwg mailing list
> > ntpwg at lists.ntp.org
> > https://lists.ntp.org/mailman/listinfo/ntpwg
>
>
> --
> ------------------------------------------------------------------------
>
> *MEINBERG Funkuhren GmbH & Co. KG*
> Lange Wand 9
> D-31812 Bad Pyrmont, Germany
> Tel.: ++49 (0)5281 9309-25
> Fax: ++49 (0)5281 9309-30
> eMail: heiko.gerstung at meinberg.de <mailto:heiko.gerstung at meinberg.de>
> Internet: www.meinberg.de <http://www.meinberg.de/>
>
> ------------------------------------------------------------------------
>
> Meinberg radio clocks: 25 years of accurate time worldwide
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://lists.ntp.org/pipermail/ntpwg/attachments/20071025/22d36b93/attachment-0001.html 