Re: [STDS-802-3-EPOC] [802.3_EPOC] Synchronization Eval Criteria for EPoC - Presentation for Jan. 8 Eval Criteria Ad Hoc Call
Bill,
All respect to Geoff, but I believe there is some disconnect between what
802.3bf does and what you describe below. I am CCing former Chair of 802.3bf
Task Force to correct anything I might misrepresent.
First, 802.3bf provides independent registers for transmit and receive paths
(see slide 16 and 17 in
http://grouper.ieee.org/groups/802/3/epoc/public/mar12/hajduczenia_01_0312.p
df), allowing you to retrieve information on min/max delay in these
directions independently. That was geared towards P2MP systems from day one
(and some P2P as well), to accommodate burst mode transmission in EPON, as
well as any non-symmetric queuing you might have in any station. Unless
there is something we did not think of at the time, I'd think P2MP systems
are very much covered by 802.3bf. There is no reason that I see why it would
not work well in EPoC.
Second, 802.3bf is not limited to 1588v2, but works with 802.1AS (see slide
8 in that deck). In fact, it was first conceived as hardware support for
802.1AS (please look at the PAR and Task Force name -
http://www.ieee802.org/3/bf/), but we later became aware of its applications
go beyond 802.1AS, and in fact can be used not only for synchronization and
time transfer purposes. That fact was reflected in the text of current
Clause 90.
Third, the method described in Clause 13 of 802.1AS relies on your 802.1AS
agent knowing (quite magically at the time) the residency time in the PHY.
802.3bf was not done at the time when 802.1AS was completed, and as such, it
was not referenced in there. However, 802.3bf provides the way to remove
magic from the picture, allowing the 802.1AS agent to actually calculate how
much time it takes for a frame to propagate through PHY in transmit and
receive directions and the reference timeframe for calculations (the
now-famous transfer indication signals - see TS_TX.indication and
TS_RX.indication on slide 10 in the deck), giving the last missing piece of
information to properly account for residency time in the given station. The
rest of the process is defined in 802.1AS.
Last, but not least, when we talk about any budget, I assume you're
suggesting an NNI-to-UNI budget here, which immediately bears the question -
how much of this budget we have for EPoC PHY to consume? Sticking an E2E
number (from our perspective NNI-to-UNI is E2E) for a device comprising much
more than 802.3 layers is not really helpful. Do we consume all of this
budget in the PHY? It is the very same problem with the E2E delay we have
been discussing for quite some time by now. People want to see delay number
for EPoC and forget that we cannot normatively nail it down. We can provide
delay budget for PHY but it will not tell you what delay your application
will experience in the future, since there are so many more factors in play
at that level.
To conclude: I think looking at this particular application for EPoC is
valuable, and we should assess the parameters you brought in when we have
PHY to work on, but I am very uncomfortable picking numbers for E2E link
right now and making them goals, when so much is up in the air right now and
we control only a small portion of it.
Marek
From: Bill Powell [mailto:bill.powell@xxxxxxxxxxxxxxxxxx]
Sent: Wednesday, January 09, 2013 00:28
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-EPOC] [802.3_EPOC] Synchronization Eval Criteria
for EPoC - Presentation for Jan. 8 Eval Criteria Ad Hoc Call
Marek,
Yes, you are correct that 802.3bf does not include a mechanism for
timestamping packets. My understanding of the usefulness of 802.3bf (after
re-reading an email from Geoff Garner) was for use over full duplex Ethernet
where the time transport protocol was IEEE1588. This could possibly also be
used as a method of time synchronization of CNUs to the EPoC CLT but might
need additional study for use in the EPoC P2MP environment.
What I've included as an example of a time transport method between the EPoC
CLT and CNUs in my presentation (on slide 9) was a method similar to the
method implemented for EPON (MPCP counter synchronization of OLT and ONU,
and use of 802.1as Clause 13). This method does not rely on the use of
802.3bf or the timestamping of the1588 messages. Since this OLT-to-ONU
synchronization method was already defined for EPON , it was also utilized
as part of the time transport protocol defined in 802.1as Clause 13 for the
media-dependent EPON time transport method.
The main point of the presentation was to look at the overall picture and
try to suggest a practical time transport error budget for EPoC to guide our
choice of time transfer protocols relative to our EPoC MAC-PHY decisions.
It was not to propose a specific solution at this point.
Regards,
Bill
-------- Original Message --------
Subject:
Re: [STDS-802-3-EPOC] [802.3_EPOC] Synchronization Eval Criteria for EPoC -
Presentation for Jan. 8 Eval Criteria Ad Hoc Call
Date:
Tue, 8 Jan 2013 23:24:31 +0000
From:
Marek Hajduczenia <mailto:marek.hajduczenia@xxxxxx>
<marek.hajduczenia@xxxxxx>
Reply-To:
Marek Hajduczenia <mailto:marek.hajduczenia@xxxxxx>
<marek.hajduczenia@xxxxxx>
To:
<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
<STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Thank you David,
For those interested in more details, please look at the presentation on
.3bf given already to this TF in March 2012 (link:
http://grouper.ieee.org/groups/802/3/epoc/public/mar12/hajduczenia_01_0312.p
df). Note the persistent lack of the word "time stamp" in this document as
well as in 802.3-2012, Clause 90. We do not do any time stamping in 802.3bf.
Marek
-----Original Message-----
From: Law, David [mailto:dlaw@xxxxxx]
Sent: Tuesday, January 08, 2013 23:14
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-EPOC] [802.3_EPOC] Synchronization Eval Criteria
for EPoC - Presentation for Jan. 8 Eval Criteria Ad Hoc Call
Hi Marek,
I believe that your description of the operation IEEE Std 802.3bf-2011 -
which now can be found in Clause 90 'Ethernet support for time
synchronization protocols' of IEEE Std 802.3-2012 - is entirely correct. As
stated in subclause 90.2 'Overview', 'The goal of this clause is to provide
an accurate indication of the transmission and reception initiation times of
all packets, as required to support various time synchronization protocols,
e.g., IEEE Std 1588-2008, and IEEE Std 802.1AS.'.
As stated in subclause IEEE Std 802.3-2012 subclause 90.7 'Data delay
measurement', 'The transmit path data delay is measured from the input of
the beginning of the SFD at the xMII to its presentation by the PHY to the
MDI. The receive path data delay is measured from the input of the beginning
of the SFD at the MDI to its presentation by the PHY to the xMII.'.
PHY registers provided maximum transmit path data delay, minimum transmit
path data delay, maximum receive path data delay and minimum receive path
data delay (for example see IEEE Std 802.3-2012 subclauses 45.2.1.104
through 45.2.1.106) which allow the calculation of the transmit and receive
path data delay variation. The critical value is this delay variation - not
the absolute delays. The greater the variation - these lower the accuracy of
the time synchronization.
Any proposal - such as the proposals we were talking about on today's call -
will have to be evaluated in respect to transmit and receive path data delay
variation to determine the accuracy of the time synchronization it will
provide.
Best regards,
David
-----
From: Marek Hajduczenia [mailto:marek.hajduczenia@xxxxxx]
Sent: 08 January 2013 22:05
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-EPOC] [802.3_EPOC] Synchronization Eval Criteria
for EPoC - Presentation for Jan. 8 Eval Criteria Ad Hoc Call
Bill,
If I recall correctly, 802.3bf does not deal with time stamping, especially
not with MPCP time stamping. All it does it provides indication of passage
of packets through the xMII, which can be then correlated with transmission
of selected packets across PHY. That combined with information on PHY delay
stored in 802.3bf registers provides enough information to calculate precise
residence time for the station.
So, I am confused what is really meant by "using a timestamping mechanism
for MPCP packets over EPoC similar to the method described in both 802.3bf
and 802.1as Clause 13". I think there is some misunderstanding as to what
802.3bf provides and misconception that it is a specification competitive to
802.1as (which it is not, it complements 802.1as as much as 1588v2).
Marek
From: Bill Powell [mailto:bill.powell@xxxxxxxxxxxxxxxxxx]
Sent: Tuesday, January 08, 2013 21:40
To: Marek Hajduczenia
Cc: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_EPOC] Synchronization Eval Criteria for EPoC -
Presentation for Jan. 8 Eval Criteria Ad Hoc Call
Marek,
Thanks for the review and comments. I agree, we have not proven yet that we
can meet this criteria, especially the time transfer error criteria on slide
17. However, due to industry requirements for MBH, and competing
technologies (GPON) that can provide even lower time transfer error
performance, I think we need to be aiming at architecting EPoC to try to
meet the goals in this presentation. Having a goal in this area will help
us when we start comparing potential implementations.
Note that on Slide 17 the time transfer error performance is labeled as
tentative, which in my mind would depend on physical limitations of what we
do in the MAC<->PHY layer for EPoC.
I agree that we will need to use a timestamping method for MPCP packets
described in 802.3bf. I discussed the differences between 802.3bf and
802.1as (the protocol for MPCP counter synchronization between the EPON OLT
and EPON ONU) with Geoff Garner at our last meeting in San Antonio. Geoff
mentioned that the timestamping method in 802.1as Clause 13 was essentially
similar to the MPCP timestamping criteria in 802.3bf, and the reason that
802.1as did not refer to 802.3bf directly is that 802.3bf was not completed
when 802.1as was balloted.
So, the intent here is to propose using a timestamping mechanism for MPCP
packets over EPoC similar to the method described in both 802.3bf and
802.1as Clause 13. I intentionally left off this level of
802.1as-vs-802.3bf detail in this presentation, as I thought it would not be
beneficial to the overall goal of:
(1) presenting a system level analysis relative to current MBH and CES
timing requirements, and
(2) propose specific Eval criteria for EPoC to support these requirements.
Regards,
Bill
-------- Original Message --------
Subject:
RE: [802.3_EPOC] Synchronization Eval Criteria for EPoC - Presentation for
Jan. 8 Eval Criteria Ad Hoc Call
Date:
Tue, 8 Jan 2013 21:13:46 +0000
From:
Marek Hajduczenia <mailto:marek.hajduczenia@xxxxxx>
<marek.hajduczenia@xxxxxx>
To:
'Bill Powell' <mailto:bill.powell@xxxxxxxxxxxxxxxxxx>
<bill.powell@xxxxxxxxxxxxxxxxxx>,
<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
<STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Bill,
My two cents on that would be simple - a valuable contribution, but adopting
such requirements without even knowing if they are achievable, is at least
premature. We do not have any working models that would allow to estimate
whether such stringent requirements are achievable.
Furthermore, despite multiple indications from my side, the support for IEEE
Std 802.3bf is being still ignored. Without it, I do not know how you can
even think of getting to the level of precision you're suggesting. The delay
through each layer in the EPoC PHY needs to be known if we can ever hope to
get 1588v2 to the precision level you're showing.
Regards
Marek
-----Original Message-----
From: Bill Powell [mailto:bill.powell@xxxxxxxxxxxxxxxxxx]
Sent: Tuesday, January 08, 2013 20:30
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: [802.3_EPOC] Synchronization Eval Criteria for EPoC - Presentation
for Jan. 8 Eval Criteria Ad Hoc Call
Hello All,
I've attached a presentation that I plan to present during the Eval Criteria
ad hoc call tomorrow (Jan. 9). Some of you may have seen most of this
presentation in a different forum, but I'm bringing this into 802.3bn
through the Eval criteria Ad Hoc for further discussion, with a recommended
performance limit for time and frequency transfer error over the EPoC link
on Slide 17 (new slide).
If the group thinks that it would be useful to present to the whole 802.3bn
group, I can also submit it for presentation for the upcoming Phoenix or a
future 802.3bn meeting.
Regards,
Bill
--
Bill Powell
Alcatel-Lucent
Fixed Access Systems Engineering
2301 Sugar Bush Road
Raleigh, NC 27612
bill.powell@xxxxxxxxxxxxxxxxxx
(O) 919-850-6462
(Cell) 919-614-3225
________________________________________________________________________
To unsubscribe from the STDS-802-3-EPOC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-EPOC
<https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-EPOC&A=1> &A=1
________________________________________
<="" p="">
________________________________________________________________________
To unsubscribe from the STDS-802-3-EPOC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-EPOC
<https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-EPOC&A=1> &A=1
________________________________________________________________________
To unsubscribe from the STDS-802-3-EPOC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-EPOC
<https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-EPOC&A=1> &A=1
_____
________________________________________________________________________
To unsubscribe from the STDS-802-3-EPOC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-EPOC&A=1