Re: [TICTOC] comments on draft-ietf-tictoc-1588overmpls-04.txt

[Stefano Ruffini <[e-mail]>]

Hi

I just saw the latest version of the draft.

I am fine with the  generic text proposed in this draft. 

Thanks
stefano

-----Original Message-----
From: Stefano Ruffini 
Sent: giovedì 6 giugno 2013 08:39
To: 'Meyer, Peter'; Shahram Davari; [e-mail]
Subject: RE: [TICTOC] comments on draft-ietf-tictoc-1588overmpls-04.txt

Hi Peter,

The point is that while technically it might be possible, if compared with a TC, the use of BC in a multi operator context looks over complicated  (note: I am not referring to a BC connected to multiple PTP domains and that in the end must just select one of them; in this case the BC would have to actually maintain different time scales, independent datasets, run independent BMCA, etc.). This can be an issue in case where  the number of clients a provider network needs to serve is significantly high.  
At least a note with some warning could be beneficial (e.g. "Note: the applicability of a BC in the context of multiple customers may need further investigation. ")

Br
stefano

-----Original Message-----
From: Meyer, Peter [mailto:[e-mail]] 
Sent: mercoledì 5 giugno 2013 21:21
To: Stefano Ruffini; Shahram Davari; [e-mail]
Subject: RE: [TICTOC] comments on draft-ietf-tictoc-1588overmpls-04.txt


Hi Stefano,

I see Figure 3 and Figure 4 apply to the text above the figures, unrelated to multiple domains.

I see the multiple domain paragraph as a separate topic, perhaps could be its own sub-section.  We have seen systems that support multiple domains either as BC or TC.  I could support removal of the mention of multiple operator domains in total, if that is agreeable.  Otherwise I see BC and TC as both valid solutions available in the market today.

Regards,
Peter


-----Original Message-----
From: Stefano Ruffini [mailto:[e-mail]] 
Sent: June 5, 2013 2:31 PM
To: Meyer, Peter; Shahram Davari; [e-mail]
Subject: RE: [TICTOC] comments on draft-ietf-tictoc-1588overmpls-04.txt

Hi Shahram, Peter

There is one point I would disagree with (I am referring to Peter's latest proposal, but the same comment would apply to the previous text):
"
An MPLS domain may serve multiple customers.  In these cases the
   MPLS domain (maintained by a service provider) may provide timing
   services to multiple customers, each having their own Timing domain.
   If so, then the LER BCs and TCs may interact with multiple grandmasters,
   and consequently multiple time references.  Also, LER/LSR TCs MUST be
   synchronized to the same master clock (which is the service provider
   Master clock), which implies that they need to choose one master to
   which it will synchronize.
"

My understanding of this statement is that a LER would need to implement multiple BC instances (with multiple clocks); this looks too complex.

The case of multiple operators, seems to me only practically applicable to figure 3 (LER TC).

I would then propose to remove the reference to LER BC in this statement .

Best regards
stefano


-----Original Message-----
From: [e-mail] [mailto:[e-mail]] On Behalf Of Meyer, Peter
Sent: mercoledì 5 giugno 2013 20:01
To: Shahram Davari; [e-mail]
Subject: Re: [TICTOC] comments on draft-ietf-tictoc-1588overmpls-04.txt

Hi Shahram,

I hope in-line with my comments I have edited the text of draft -04.

I'm not sure how the formatting worked out (PC-based, line length, page length).

I modified sections 3, 4, 18, 19 & 20.

I include here both original and modified version so you can diff (most of the changes in 3 are due to line length / re-org, not actual text changes).

File attachment:
"draft-ietf-tictoc-1588overmpls-04-org.txt" (52 KB) "draft-ietf-tictoc-1588overmpls-04-pm.txt" (53 KB)


Regards,
Peter


-----Original Message-----
From: Shahram Davari [mailto:[e-mail]]
Sent: June 4, 2013 7:40 PM
To: Meyer, Peter; [e-mail]
Subject: RE: [TICTOC] comments on draft-ietf-tictoc-1588overmpls-04.txt

Hi Peter,

Thanks for your god comments. Please see my responses inline. Please let me know whether you are satisfied with my responses since I am updating the draft in the next couple of days.

Thanks
Shahram

-----Original Message-----
From: [e-mail] [mailto:[e-mail]] On Behalf Of Meyer, Peter
Sent: Monday, March 11, 2013 5:49 PM
To: [e-mail]
Subject: Re: [TICTOC] comments on draft-ietf-tictoc-1588overmpls-04.txt

Hi Shahram et al,

Some comments on draft -04.


1) Section 3.  "A generic method is defined in this document that does not require deep packet inspection at line rate, and can deterministically identify Timing messages.  The generic method is applicable to MPLS and MPLS-TP networks."

May want to add that this would apply only to one-step TC (I imagine that is the point of the correction at line rate).  A two-step TC would need to do deep packet inspection as it uses sourcePortIdentity & sequenceId fields for Follow_up or Delay_Resp.

SD> For 2-step, all that is needed is for an interface to detect the
Timing message (based on top MPLS Label)  and record the Time along with some information from the Timing packet and send them to CPU while forwarding the Timing message as usual. Then the CPU based on the information received generates Follow-up message and injects it in to the LSP.  Therefore it is the job of CPU to look at sourcePortIdentity & sequenceId fields not the Hardware.


2) Section 4.  "An MPLS domain can serve multiple customers.  This means that the MPLS domain (maintained by a service provider) may provide timing services to multiple customers, each having their own Timing domain.  Therefore LER BCs need to interact with multiple grandmasters, and consequently multiple time references."

This should be re-phrased.  It switches from an optional situation ("may provide timing services") to mandatory situation ("LER BCs need to
interact").   Some words such as "in such a deployment scenario, ...."
and replace "can" with "may".   We have seen at ITU at least (with
participation from operators BT, FT, DT, CMCC, AT&T, Sprint, etc.) that this multiple operator domain case was not useful enough to be included in standardization process for Telecom networks.

SD> I rephrased it as following:

" An MPLS domain can serve multiple customers.  This means that the
   MPLS domain (maintained by a service provider) may provide timing
   services to multiple customers, each having their own Timing domain.
   Therefore in such deployment scenarios, LER BCs may interact with multiple grandmasters, and consequently multiple time references.  Also, LER/LSR TCs MUST be synchronized to the same master clock (which is the service provider Master clock), which implies that they need to choose one master to synchronize to."


3) Section 19.  "For transporting such peer delay measurement messages a single-hop LSP SHOULD to be created between the two adjacent LSRs engaged in peer delay measurement to carry peer delay measurement messages. Other methods such as PTP transport over Ethernet MAY be used for transporting peer delay measurement messages if the link between the two routers is Ethernet."


This new statement to handle peer-delay (which in earlier drafts did not have a communication path listed), also allows a BC to be embedded in an LSR with a communication path between BC's as a single-hop LSP (mentioned March 22, 2012 in my feedback to draft -03, subject "[TICTOC] Updated 1588 over MPLS draf-03").

SD> When a single-hop LSP is created then I call the both ends of that
LSP, LERs and not LSRs.
Note that a router can act as LSR for some LSPs and LER for other LSPs.
The point is that we can't
Terminate a packet in the middle of an LSP (i.e., in an LSR). So BC functionality can't exist in the middle Of a PTP LSP. So I think it is a matter of terminology otherwise I agree that you could have BC in the core (not edge). 

Architecture diagrams Figure 1 and Figure should be updated to reflect
the possibility of either BC or TC implementation.   Section 18.2 and
section 18.3 and section 21 should also be updated to reflect the possibility of either BC or TC.  I understand the RFC is intended to be generic and not targeted only at TC.



4) Section 20.  "When the MPLS network (provider network) serves multiple customers, it is important to maintain and process each customers clock and Timing messages separately from other customers to
ensure there is no cross- customer effect.   For example if an LER BC is
synchronized to a specific grandmaster, belonging to customer A, then the LER MUST use that BC clock only for customer A to ensure that customer A cannot attack other customers by manipulating its time."

This seems much more applicable to the TC LSR and should be stated.
>From section 4 we see the TC uses the primary synchronization domain (that of the service provide) to correct PTP messages.

SD> I don't see how in a TC LSR, one customer timing can affect other
customer timings?
Since in a TC, we just update the Timestamp of a packet based on the Service provider Time.
May be I have not understood your point. Could you please clarify with an example.



5) Section 20.  "Timing messages (as opposed to regular customer data) SHOULD not be encrypted or authenticated on an end-to-end basis."  I think there is a security draft in parallel being developed that may be relevant to that statement.

SD> I have updated the  text as following:

" Timing messages MAY be encrypted or authenticated, provided that the LERs/LSRs that are Timing capable/aware can authenticate/ decrypt the timing messages."




Regards,
Peter


-----Original Message-----
From: [e-mail] [mailto:[e-mail]] On Behalf Of [e-mail]
Sent: February 23, 2013 2:23 AM
To: [e-mail]
Cc: [e-mail]
Subject: [TICTOC] I-D Action: draft-ietf-tictoc-1588overmpls-04.txt


A New Internet-Draft is available from the on-line Internet-Drafts directories.
 This draft is a work item of the Timing over IP Connection and Transfer of Clock Working Group of the IETF.

	Title           : Transporting Timing messages over MPLS
Networks
	Author(s)       : Shahram Davari
                          Amit Oren
                          Manav Bhatia
                          Peter Roberts
                          Laurent Montini
	Filename        : draft-ietf-tictoc-1588overmpls-04.txt
	Pages           : 36
	Date            : 2013-02-22

Abstract:
   This document defines the method for transporting Timing messages
   such as PTP and NTP over an MPLS network.  The method allows for the
   easy identification of these PDUs at the port level to allow for port
   level processing of these PDUs in both LERs and LSRs.

   The basic idea is to transport Timing messages inside dedicated MPLS
   LSPs.  These LSPs only carry timing messages and possibly Control and
   Management packets, but they do not carry customer traffic.

   Two methods for transporting Timing messages over MPLS are defined.
   The first method is to transport Timing messages directly over the
   dedicated MPLS LSP via UDP/IP encapsulation, which is suitable for
   MPLS networks.  The second method is to transport Timing messages
   inside a PW via Ethernet encapsulation, which is suitable for both
   MPLS and MPLS-TP networks.


The IETF datatracker status page for this draft is:
https://datatracker.ietf.org/doc/draft-ietf-tictoc-1588overmpls

There's also a htmlized version available at:
http://tools.ietf.org/html/draft-ietf-tictoc-1588overmpls-04

A diff from the previous version is available at:
http://www.ietf.org/rfcdiff?url2=draft-ietf-tictoc-1588overmpls-04


Internet-Drafts are also available by anonymous FTP at:
ftp://ftp.ietf.org/internet-drafts/

_______________________________________________
TICTOC mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/tictoc


_______________________________________________
TICTOC mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/tictoc