Stewart, Shahram and all,
Here is my proposal based on reading this thread and some thinking.
Instead of trying to specify the mandatory default for the timestamp format,
let's mandate something else:
1. Ability to convert the timestamps from PTP to NTP for 1-way delay
measurements.
* The computational load associated with this operation could result in
relatively low rate of 1-way DM messages (say 1 pps?) in line with the general
direction of the draft which says that loss and delay measurement should not
stretch the network resources.
* The direction (PTP to NTP) of conversion has been chosen because
computation of differences is simpler for NTP timestamps.
2. Ability to compute differences between pairs of timestamps in the same
format both for NTP and PTP timestamps for 2-way delay measurements. This
operation does not require conversion from one format into another.
Neither of the two items above carries with it any special HW requirements.
Hence, IMHO this approach, if accepted, would promote backward compatibility
with the installed LSR base while allowing new equipment to benefit from the
HW-based PTP-oriented timestamping mechanisms.
Below you can find the chain of reasoning that has led me to this proposal
(you may skip this part if you are not interested:-).
1. I tend to agree with Stewart: neither rollover nor leap seconds are really
relevant for the delay measurement application. (I did not think so
yesterday:-().
2. AFAIK (and I have discussed in a personal email exchange with the authors
of the draft), the required accuracy of delay measurements is not specified -
neither in the draft itself (intentionally) nor in other relevant standards.
* This thread really confirms that - the best I see is a somewhat vague
reference to "expectations of the Transport network operators". Did I miss
something here?
* Without these requirements references to comparative "accuracy of the
timestamps" for NTP vs. PTP in this thread do not make too much sense to me
3. The available accuracy of delay measurements depends on many factors in
addition to the timestamp resolution. E.g., we all know that one-way delay
measurements require ToD synchronization between two endpoints. It is also
widely known that, e.g., TDD-based mobile application require ~1 microsecond
accuracy of ToD synchronization. Hence I conclude that the achievable accuracy
of 1-way delay measurements will be probably in the microseconds range, no
matter which protocol and what kind of timestamps we use (as long as the
granularity of the timestamps is in microseconds). Hence I agree with Stewart -
quite a few the bits in the nanoseconds part of the PTP timestamp can be
actually ignored
4. The draft we discuss here is NOT an MPLS-TP draft. It is a generic MPLS
draft with an accompanying MPLS-TP document. IMO imposing a TP-derived
requirement on non-TP equipment doesn't make too much sense
5. The draft inherently supports the situation when the requester and the
responder use different timestamping formats:
* For 2-way DM only ability to compute the difference between two NTP
timestamps and two PTP ones is required
* For 1-way DM conversion to a single format (e.g., NTP) is required.
While this may be computationally heavy, it is not an operation that has to be
done in HW or even in hard real-time. If, say, you send the delay measurement
packets once a second, the originator would have 1" to do this "heavy work" -
lots of cycles in a modern CPU! BTW, one of the issues *not addressed* in the
draft is the rate at which delay measurement messages are supposed to be sent
* The installed base of LSRs will probably benefit from using
SW-generated NTP timestamps (with whatever accuracy can be achieved)
* New equipment that supports HW timestamping for PTP "close to PHY" will
benefit from using PTP timestamps.
My 2c,
Sasha
________________________________
From: [email protected] [[email protected]] On Behalf Of Stewart Bryant
[[email protected]]
Sent: Friday, March 04, 2011 12:48 AM
To: Shahram Davari
Cc: '[email protected]'; '[email protected]'; '[email protected]';
'[email protected]'; '[email protected]'
Subject: Re: [mpls] [TICTOC] draft-ietf-mpls-loss-delay Timestamp
I do not think that rollover or leap seconds are much of a problem in this
application. We are measuring the transit time of a packet across a network.
That is less than a second. Rollover or leap second errors mean that we make a
mistake once in every 136 years in the first case and maybe never in the case
of leap seconds. I am sure that it would be acceptable to simply state that for
that for that one second at some time in the future we cannot make that
measurement.
BTW we are only going to use 32 bits of seconds and 32 bits of sub seconds in
the 1588 case, and even then most of the sub second bits will be ignored. I
don't thing that we need to be measuring time of flight over 8" of cable.
Stewart
On 03/03/2011 22:12, Shahram Davari wrote:
Agree.
Shahram
________________________________
From: Ron Cohen <[email protected]><mailto:[email protected]>
To: Shahram Davari
Cc: [email protected]<mailto:[email protected]>
<[email protected]><mailto:[email protected]>;
[email protected]<mailto:[email protected]> <[email protected]><mailto:[email protected]>;
[email protected]<mailto:[email protected]>
<[email protected]><mailto:[email protected]>;
[email protected]<mailto:[email protected]>
<[email protected]><mailto:[email protected]>
Sent: Thu Mar 03 13:35:11 2011
Subject: Re: [TICTOC] draft-ietf-mpls-loss-delay Timestamp
Hi Shahram, I guess adding a requirement to take timestamp rollover into
account would do. If there is a strong reason to also leave NTP timestamp as an
option, a note explaining the ramification of leap seconds should be added. I
suggest that a continuous timescale (i.e. PTP) would be selected in the draft
as the default/must option. Best, Ron
On Thu, Mar 3, 2011 at 9:45 PM, Shahram Davari
<[email protected]<mailto:[email protected]>> wrote:
Hi Ron,
Although the PTPv2 is 80 bytes, but I don’t see a need to use all 80 bytes for
delay measurement. The upper 2^32 seconds covers almost 136 years, which should
be good enough to measure delay. Also lets’ be compatible with Y.1731 format
since that HW already exist.
Thx
Shahram
From: Ron Cohen [mailto:[email protected]<mailto:[email protected]>]
Sent: Thursday, March 03, 2011 11:22 AM
To: [email protected]<mailto:[email protected]>
Cc: Shahram Davari; [email protected]<mailto:[email protected]>;
[email protected]<mailto:[email protected]>;
[email protected]<mailto:[email protected]>
Subject: Re: [TICTOC] draft-ietf-mpls-loss-delay Timestamp
Hi,
I apologize in advance if my comments below where already been discussed.
The draft specify using PTPv1 timestamp format and not PTPv2 timestamp format.
The seconds part of the PTPv2 timestamp is 48bits, and not 32bits as written in
the draft. NTP timestamps will roll over in 2036. PTPv2 timestamps do not
rollover.
I don't think its advisable to include timestamps that rolls over ~20 years
from now.
In addition, NTP timestamp, unlike PTP timestamp, 'jumps' in leap seconds
events. This makes calculation of time differences hard, and would render some
measurements ambiguous. It is much easier to use continuous timescale like PTP.
In my opinion changing the timestamp format to PTPv2 timestamp format (i.e.
48bit/32bit sec/ns) should be the right way to go. I would also remove the NTP
timestamp format option to avoid leap seconds confusion and 2036 rollover
events.
Best,
Ron
On Thu, Mar 3, 2011 at 8:26 PM,
<[email protected]<mailto:[email protected]>> wrote:
My feeling is 1588 is the right answer in the long run since it is already
coupled to the hardware and widely deployed in networks using the "packet
clock derived time" in precise time and frequency delivery.
Pat
"Shahram Davari"
<davari@broadcom.
com> To
Sent by:
"[email protected]<mailto:[email protected]>"
tictoc-bounces@ie
<[email protected]<mailto:[email protected]>>,
tf.org<http://tf.org>
"[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>>
cc
"[email protected]<mailto:[email protected]>"
<[email protected]<mailto:[email protected]>>
03/03/2011 10:05 Subject
AM Re: [TICTOC]
draft-ietf-mpls-loss-delay
Timestamp
Hi Stewart,
Accurate delay measurement requires the Timestamp to be done in HW. It is
true that most routers support NTP today, but the majority of those are
really maintained in Software and AFAIK most HW based timestamps
implementations are 1588 based and really the industry is shifting toward
1588 and SyncE for network synchronization. So I support 1588 as being the
default.
Thanks,
Shahram
From: [email protected]<mailto:[email protected]>
[mailto:[email protected]<mailto:[email protected]>] On Behalf Of
Stewart Bryant
Sent: Thursday, March 03, 2011 8:07 AM
To: [email protected]<mailto:[email protected]>
Cc: [email protected]<mailto:[email protected]>
Subject: [TICTOC] draft-ietf-mpls-loss-delay Timestamp
We have received a LC comment on draft-ietf-mpls-loss-delay concerning the
default timestamp.
In the first version of the draft we proposed NTP, but following initial
comments from the MPLS-TP community we changed to IEEE1588. This
requirement for IEEE1588 can be traced back to the choice of using IEEE1588
in Y.1731.
We have now received a LC request to change the default back to NTP.
NTP is the "natural" choice for an IETF protocol. NTP is specified in IETF
and is used in other MPLS protocols such as LSP ping. It is implemented on
almost every host and every router. However in general the implementations
provides a relatively low precision timestamp, and the NTP time
distribution infrastructure operates on a best effort basis. Thus even a
good client implementation would normally have a relatively low quality
path to the server, which would result in a low quality of timestamp. NTP
could be made to work to higher accuracy, but defacto upgrading NTP and
providing high quality NTP paths to the time servers is not getting much
attention. Computing timestamp differences is easier with NTP.
On the other hand IEEE1588 has defacto become the two way time tranfer
protocol for precision applications. IEEE1588 only provides high quality
time in well engineered networks with some form of hop by hop assistance.
It is not widely implemented other than in equipment targeted to specific
markets, although one of those markets is in mobile backhaul applications
where we expect to see significant initial deployment of MPLS-TP. Computing
timestamp differences is harder with IEEE1588
The loss-delay work is targeting to be able to do one way packet delay
measurement, so it does need a higher quality of timestamp than is required
in general purpose network instrumentation.
Converting from IEEE1588 to NTP is not trivial, since they use different
epochs, and different representations of sub-second time. In a
"traditional" NTP implementation the time error due to on the fly
conversion is likely to be small compared to the time error in the time
synchronization system, but an IEEE1588 system would need hardware
conversion to maintain the accuracy for an NTP timestamp.
So the question arises, should we make IEEE1588 or NTP the default for
draft-ietf-mpls-loss-delay. Much as I would like to suggest that we should
go back to NTP for consistency with other IETF protocols, I have difficulty
reconciling this with the situation in network deployments and thus suggest
that we continue to use IEEE1588.
What is the opinion of the working group?
Stewart (speaking as a draft editor)
_______________________________________________
TICTOC mailing list
[email protected]<mailto:[email protected]>
https://www.ietf.org/mailman/listinfo/tictoc
_______________________________________________
TICTOC mailing list
[email protected]<mailto:[email protected]>
https://www.ietf.org/mailman/listinfo/tictoc
_______________________________________________
mpls mailing list
[email protected]<mailto:[email protected]>
https://www.ietf.org/mailman/listinfo/mpls
--
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/index.html
_______________________________________________
TICTOC mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/tictoc
