Re: [PWE3] draft-manhoudt-pwe3-tsop-00.txt and IP

[Alexander Vainshtein <[e-mail]>]

Gert, and all,
Please see some comments inline below.

Regards,
     Sasha

> -----Original Message-----
> From: Gert Manhoudt [mailto:[e-mail]]
> Sent: Wednesday, July 18, 2012 11:33 AM
> To: Alexander Vainshtein; [e-mail]; Yaakov Stein
> ([e-mail])
> Cc: [e-mail]; [e-mail];
> pwe3; Ron Cohen
> Subject: Re: [PWE3] draft-manhoudt-pwe3-tsop-00.txt and IP
> 
> Stewart, Yaacov and Alexander,
> 
> Thanks for your comments on draft-manhoudt-pwe3-tsop-00. It is our
> intent to request the PWE3 WG to accept this document as a working group
> draft. 
 [[[Sasha]]] I may be running ahead of things, but I would not support adoption of the draft - at least in its present form.

> At this stage I want to address those points that could be
> interpreted to  imply that the TSoP approach as such is not acceptable.
> The other comments are of a less principal nature and can be addressed
> later (and can probably be fixed by adapting the text).
> 
> 1. TSoP does not require the termination of any STM-N overhead:
> Yaakov's e-mail contains the statement: "this function is at least a
> section (in SDH terminology a regenerator section) termination". I
> believe that this statement is not true: There is no reason why STM-N
> signals can not be carried transparently over some server layer with
> minimal interaction between client and server (as should always be the
> goal).
> 
> The correct architectural analogue would be to compare a TSoP PW section
> to an OTN network that transports an STM-N client signal. The mapping of
> an STM-16/OC-48 in an ODU1 does not require any processing of STM-N
> overhead or even framing for alignment purposes: All STM-16 bits are
> directly written into the OPU1 payload area. This is described in G.709
> section 17.2.1. In terms of functional blocks this is described in G.798
> section 14.3.1. Note that Appendix I.1 of this same Recommendation
> describes the mapping of STM-N in OPU-k using OPTIONAL RS-monitoring,
> just like TSoP.
> 
[[[Sasha]]] Do you imply that you TSoP carries STM-N unscrambled?
>
> Because a TSoP PW section (STM-N over PSN) is modeled to behave itself
> in the same way as an STM-N over OTN section, the TSoP decapsulation
> function inserts Generic AIS as defined in G.709, rather than SDH
> multiplex section AIS as defined in G.707, in case the client signal is
> interrupted.
[[[Sasha]]] To the best of my understanding:
(a) You cannot insert MS-AIS (because you do not know where the MS is)
(b) Any standard SDH framer would recognize the Generic AIS as RS-LOF
(b) inserting "all ones" would actually have the same effect.
> 
> As a side remark: Yaakov's mail also contains the statement: "(since it
> does not divide up STM-Ns, it is not a line termination)". Note that
> this directly contradicts the remark that the S1-byte must be processed.
> If that were really necessary, it would not be sufficient to limit the
> required SDH functionality to a regenerator section termination, but
> rather a full multiplex section termination (SONET: line termination)
> would be necessary, since the S1-byte is located in the MS part of the
> STM-N overhead and not in the RS part: SDH regenerators pass the S1-byte
> transparently.
> 
> 2. TSoP passes the S1-byte transparently:
> The S1-byte in the STM-N multiplex section overhead carries the
> "traceability information" of the STM-N signal, i.e. the quality level
> of the master clock at the beginning of a chain of slaved STM-N systems.
> Normally this quality level is PRC/PRS, but during fault situations, it
> may lower to SSU/ST2 or even SEC/ST3. Downstream STM-N systems can use
> this information to select another reference signal or to go in
> hold-over mode.
> 
> The question is now whether the passage of an STM-N signal over a
> TSoP/PSN section should affect the value of the S1-byte. We believe that
> this is not the case: The nature of a passage of an STM-N signal through
> a TSoP Pseudowire is such that the traceability does not change, since
> the (long term average of the) frequency of the STM-N signals before
> entering the TSoP PW and after leaving the TSoP PW is the same: No bits
> are created or deleted from the STM-N client signal. This means that if
> the long term frequency accuracy of an STM-N signal is 1E-11 (i.e. it is
> traceable to PRC/PRS), that this will not change by passing the TSoP PW,
> i.e. the traceability is invariant.
> 
[[[Sasha]]] I think that your interpretation of S1 is inaccurate.
In SDH (and SONET) it carries the information about the traceability level of the clock that has been used to generate the current SDH bit-stream.  The TSoP endpoint should do exactly the same IMO.

So what is the quality of the clock that the TSoP endpoint uses? There are at least three factors that affect this clock:
(a) Characteristics of the local oscillator used for clock recovery
(b) Quality of the original clock
(c) Specifics of the clock recovery algorithm. 

(c) above may be not all that relevant with Differential clock recovery that you propose to use for TSoP. For Adaptive clock recovery its impact on wander of the recovered SAToP and CESoPSN  clock is known to be critical - and this in spite of the fact that these protocols do not add or remove any bits!

(a) and (b) are not really under your control, but I would guess that the actual quality of the recovered clock cannot be higher than the worst of these two. So if your incoming SDH stream has been clocked by the PRC, but your common clock used for Differential clock recovery is Stratum 3E, then the quality of the recovered clock would not be better that Stratum 3E - and sending out the S1 value that reports PRC quality would be wrong.

The draft could, of course, indicate (e.g., in its applicability section) that S1 values sent out by the TSoP endpoint should not be believed. It could also recommend synchronizing the two PEs containing a pair of TSoP endpoints to a PRC-class common clock - not a very far-reaching assumption taking into account availability of GPS clock sources. But without these caveats the draft simply looks as containing technically wrong statements.
 
> Of course, during a failure, the TSoP decapsulation function substitutes
> the G-AIS signal which is no longer frequency synchronous with the input
> STM-N. In this case the downstream STM-N node will immediately detect
> Loss of Frame and will no longer be able to read the S1-byte. According
> to standard SDH reference selection procedures (see G.781), this node
> will no longer consider such a signal to be an acceptable
> synchronization reference.
> 
> 3. A TSoP PWE3 can be carried over UDP/IP:
> This option has been included in the TSoP draft purely because it is
> present in the SAToP standard and to keep the encapsulation options
> aligned. If experience has shown that transport over an UDP/IP PSN of
> higher bitrate circuit emulation PW's is not recommendable, the authors
> have no problem to remove this option.
> 
> 
> Regards,
> Gert Manhoudt.
> 
> 
> 
> On 10 jul 2012 11:01, Alexander Vainshtein wrote:
> > Hi all,
> > Several comments on the draft in question:
> >
> > 1. The draft seems to follow closely RFC 4553, including support of
> IPv4/IPv6 as the PSN, congestion handling etc. I echo Stewart's question
> regarding usage of IPv4/IPv6 PSN (with limited TE capabilities) for carrying
> such high-rate CBR flows as STM-1/OC-3 and higher.
> >
> > 2. The draft correctly mentions RFC 4842 as an alternative (and approved
> by the IETF) method for emulating SONET/SDH over PSN. However, it is
> worth noting that one of the reasons this method has been preferred to
> transparent passing of the entire SONET/SDDH bit stream (e.g., as proposed
> in one of the early precursors of RFC 4553 - see
> http://www.watersprings.org/pub/id/draft-vainshtein-cesopsn-01.txt, Annex
> B) to was elimination of the need to manipulate the S1 byte in the
> SONET/SDH overhead. The lower nibble of this byte contains encoded quality
> of the system clock of the SONET/SDH mux that has been used to drive the
> outgoing SONET/SDH bit stream. Simply passing tits value transparently thru
> the PSN does not make any sense (and, indeed, potentially violates integrity
> of clock distribution). The IWF proposed in the draft cannot access this byte
> because it is unaware of the SONET/SDH boundaries, and neither can this be
> done by the  OSn_TT_So & OSn/RSn_A_So/  OSn_TT_Sk & OSn/RSn_A_Sk
> functional blocks to which the IWF is connected. IMHO the draft should
> explain in some detail how this issue is resolved. Alternatively, it should state
> that the SONET/SDH bit stream that has passed thru the TSoP PW MUST NOT
> be used as the clock source by the CE device.
> >
> > 3. The draft defines only LOS as the mandatory SONET/SDH AC fault to
> affect setting of the L bit in the PW control word. However, if Loss of Frame
> defect is encountered (dLOF),  de-scrambling of the incoming SONET/SDH
> stream cannot be completed, and the payload of the TSoP would become
> junk; hence IMO LOF must be added as a mandatory trigger for asserting the
> SF condition in Section 3.1.  (And, while the reference to the ITU-T G.783
> functional blocks implies that, it would be nice to explain that the incoming
> SONET/SDH bit stream is segmented and encapsulated after de-scrambling -
> for the sake of the readers that do not parse the ITU-T-ish easily:-)
> >
> > 4. The draft seems to imply that TSoP requires, as a pre-requisite,
> availability of common high-quality clock for the pair of PEs terminating a
> TSoP PW by both mandating the use of the minimal RTP header in the
> encapsulation and not recommending usage of adaptive clock recovery for
> these PWs. However, this important restriction is not reflected in the
> Applicability Statement section of the draft.
> >
> > 5. The draft mentions the need not to exceed the path MTU between the
> pair of PEs in the case of IPv4/IPv6 PSN but not in the case of the MPLS PSN
> (where it obviously applies).
> >
> > Hopefully these notes will be useful.
> >
> > Regards,
> >       Sasha
> >
> >> -----Original Message-----
> >> From: [e-mail] [mailto:[e-mail]] On
> Behalf
> >> Of Stewart Bryant
> >> Sent: Tuesday, July 10, 2012 10:57 AM
> >> To: pwe3
> >> Subject: [PWE3] draft-manhoudt-pwe3-tsop-00.txt and IP
> >>
> >> Do the authors think that running transparent
> >> SONET emulation over IP is likely in practice, or is
> >> this included because SAToP included it?
> >>
> >> Stewart
> >> (Speaking as an individual)
> >>
> >>
> >> _______________________________________________
> >> pwe3 mailing list
> >> [e-mail]
> >> https://www.ietf.org/mailman/listinfo/pwe3
> > This e-mail message is intended for the recipient only and contains
> information which is CONFIDENTIAL and which may be proprietary to ECI
> Telecom. If you have received this transmission in error, please inform us by
> e-mail, phone or fax, and then delete the original and all copies thereof.
> >
> 
> 


This e-mail message is intended for the recipient only and contains information which is CONFIDENTIAL and which may be proprietary to ECI Telecom. If you have received this transmission in error, please inform us by e-mail, phone or fax, and then delete the original and all copies thereof.