Re: [CCAMP] OSPF-TE extensions for G.709 OTN draft submitted: draft-ashok-ccamp-gmpls-ospf-g709-01

[Rajan Rao <[e-mail]>]

Pietro,

On Minimum-LSP-BW:    You are correct. We fully agree on setting the BW value to lowest switchable container.   There is no issue in doing so as ODUflex BW advertisement is moved to SCSI. We will clarify this in the draft. Thanks for the correction.

On Termination/Switching:    We are not able to understand the need.  It will be helpful if you could show couple of use cases with diagrams.
Is this trying to address switch blocking scenarios?

Thanks
Rajan



From: GRANDI, PIETRO VITTORIO (PIETRO VITTORIO) [mailto:[e-mail]]
Sent: Friday, October 29, 2010 2:20 AM
To: Rajan Rao
Cc: [e-mail]; Snigdho Bardalai; Ong, Lyndon; Khuzema Pithewan; Fatai Zhang; Daniele Ceccarelli; BELOTTI, SERGIO (SERGIO); Varma, Eve L (Eve); Biao Lu; Ashok Kunjidhapatham
Subject: RE: [CCAMP] OSPF-TE extensions for G.709 OTN draft submitted: draft-ashok-ccamp-gmpls-ospf-g709-01

Hello
in line and in RED to differentiate from previous statements..

Pietro


============================================
Pietro Vittorio Grandi
Terrestrial Optics Unit -  System Architecture
Alcatel-Lucent Vimercate (Italy)
Tel: [phone]
Mail: [e-mail]
============================================
Put your hand on a hot stove for a minute, and it seems like an hour.
Sit with a pretty girl for an hour, and it seems like a  minute. That's relativity.
(A. Einstein)

________________________________
From: Rajan Rao [mailto:[e-mail]]
Sent: venerdì 29 ottobre 2010 2.45
To: GRANDI, PIETRO VITTORIO (PIETRO VITTORIO)
Cc: [e-mail]; Snigdho Bardalai; Ong, Lyndon; Khuzema Pithewan; Fatai Zhang; Daniele Ceccarelli; BELOTTI, SERGIO (SERGIO); Varma, Eve L (Eve); Biao Lu; Ashok Kunjidhapatham
Subject: RE: [CCAMP] OSPF-TE extensions for G.709 OTN draft submitted: draft-ashok-ccamp-gmpls-ospf-g709-01

Pietro, et al

Please see response inline

Thanks
Rajan

From: GRANDI, PIETRO VITTORIO (PIETRO VITTORIO) [mailto:[e-mail]]
Sent: Thursday, October 28, 2010 1:31 AM
To: Rajan Rao
Cc: [e-mail]; Snigdho Bardalai; Ong, Lyndon; Khuzema Pithewan; [e-mail]; Fatai Zhang; Daniele Ceccarelli; GRANDI, PIETRO VITTORIO (PIETRO VITTORIO); BELOTTI, SERGIO (SERGIO); Varma, Eve L (Eve)
Subject: RE: [CCAMP] OSPF-TE extensions for G.709 OTN draft submitted: draft-ashok-ccamp-gmpls-ospf-g709-01

Hello all,

some considerations.

1) Draft Ashok states that

                                                 Max LSP Bandwidth

      Number of OPU TSs =      -------------------------

                                                  Min LSP Bandwidth

This means that the MIN LSP Bandwidth is always set to the nominal size of 1 TS.
So it can have only two values 1.25 and 2.5 Gbit/s.
[Rajan] {
Yes. This is correct.  But the exact value is nominal rate depending on the ODU layer involved.
}
The min LSP bandwidth for IETF should instead provide the minimum switching capability. (see RFC4202)
[Rajan] {
 We have only 2 TS variants defined for (Switching-Capability=TDM &&  Encoding=ODUk ). Not sure what the comment is about.
}
[[PG]] The comment is about the fact that Min LSP bandwidth by RFC4202 has to report the bandwidth associated to a switchable container.
By G.709 Tributary Slots  are not switchable containers, they are just a way to partition bandwidth available on a link in such a way
to have max flexibility.  Their scope is limited to the link and are never sent to the matrix. While nominal speed of the Tributary slots is dependent from the server (e.g. ODU-3 TS has different size from ODU-2 TS), the ODU speed is always the same independently from the server.  This has to be the reference for min LSP.
}

With the above relationship it seems that a NE that supporting ODU1 and higher rate containers
and exporting interfaces with TS size = 1.25 Gbit/s should always declare a Min LSP Bandwidth = to 1.25Gbit/s
that is not corresponding to a supported switching capability.
[Rajan] {
This is not true.  A link which is G.709-v1 based will advertise Min-LSP-BW=2.5G (ODU1 rate) not 1.25G (ODU0 rate). Please ref to ODU0 example in the previous chain.
[[PG]] I do not understand why you say so. The draft does not report that the relation is limited to G.709v1 interfaces. What I say happens
always when a matrix does not support ODU0 switching.  Again, the tributary slot size  by G.709 definition is not switched in matrix, and it has not to be considered as the equivalent of the ODU-1/ODU0.
}

This behavior is absolutely independent from the fact that ODU-flex is supported or not.

2)    Snip:

"Not clear what the issue is.  We have listed 2 approaches in the slides sent out yesterday. To reiterate the case where ODUflex is not supported:
 With  backward compatibility option, the ODUflex goes in to a separate sub-TLV within SCCI. If ODUflex is not supported on a link, this sub-TLV will not be included; i.e, ISCD::Max-LSP-BW is not used for ODUflex & doesn't become zero."

We would like first to observe that the ashok draft does not report this specific behavior. Secondary
we would make note that with this approach the MAX-LSP bandwidth in any case should be interpreted differently
when the ODU-flex capacity is exhausted.
If you think to a bundle link with ODU-flex capacity, then when
UDU-flex capacity is available the MAX-LSP bandwidth is used in a certain way whereas when the ODU-flex capacity is exhausted the
ODU-flex TLV is no more advertised and MAX-LSP bandwidth behavior is different.
we do not see a reason that motivates the introduction of this kind of complication in routing.

[Rajan] {
The ODUflex confusion may be due to lack of details from us.  In the ppt sent earlier we mentioned about how to handle backward compatibility & ODUflex.  The minor modifications are captured in our draft-02 version (section 5.4.2).  Please review and let us know if there are still any issues.
http://www.infinera.com/rfc/draft-ashok-ccamp-gmpls-ospf-g709-02.txt

}


3)    Snip:
Thirdly, it can not support priority flexibility, so it is not efficient from the routing scalability perspective.
[Rajan] {
The whole idea is to stay with the GMPLS architecture already in place.  The technology specific extensions should go into SCSI. This has been done successfully for SONET/SDH case (Ref to examples in RFC 4202). Do not see a reason to deviation from the original GMPLS model.
}

The original idea for GMPLS has been already discussed in the frame work document draft-ietf-ccamp-gmpls-g709-framework-03. The framework is the only
reference for requirements that have to be supported. Moreover, as required by CCAMP chair an example of the difference in efficiency between the original IETF format
and the new format is shown in the  draft draft-bccdg-ccamp-gmpls-ospf-agnostic.
[Rajan] {
We have shown in our draft that we can take care of OTN extensions in SCSI.  It doesn't make sense to have two ISCDs (ISCD+BA) for the same switching type (=TDM). Unnecessary.
}
[[PG]] This is the reason why we have created an "agnostic" format .

4)    Differentiation between termination and switching capabilities is a requirement contained in the framework document. The differentiation allows an operator
to engineer freely its network keeping into account capabilities offered by current technology. Current technology allows switching only a stage at once in the matrix.
Of course this can be overcome with very costly muxponder boards that by the way could have limitations related to which containers can be terminated and which can be
only adapted and sent to matrix for switching. In general a routing specification must not force an operator to use costly solutions (multi-stage multiplexing)
in order to operate the network. As a consequence the routing specification must provide the means to allow manage the current technology indicating clearly which are the capabilities of each interface in terms of terminations and switching.
[Rajan] {
 Responded in a separate chain on why we do not need T/S flags.
}

[[PG]] Responded in a separate chain on why this is needed.

Pietro & Sergio


============================================
Pietro Vittorio Grandi
Terrestrial Optics Unit -  System Architecture
Alcatel-Lucent Vimercate (Italy)
Tel: [phone]
Mail: [e-mail]
============================================
Put your hand on a hot stove for a minute, and it seems like an hour.
Sit with a pretty girl for an hour, and it seems like a  minute. That's relativity.
(A. Einstein)

________________________________
From: [e-mail] [mailto:[e-mail]] On Behalf Of Rajan Rao
Sent: martedì 26 ottobre 2010 21.15
To: Fatai Zhang; Daniele Ceccarelli
Cc: [e-mail]; Snigdho Bardalai; Ong, Lyndon; Khuzema Pithewan; [e-mail]
Subject: Re: [CCAMP] OSPF-TE extensions for G.709 OTN draft submitted: draft-ashok-ccamp-gmpls-ospf-g709-01

Fatai,

Please see response inline, within { }

Thanks
Rajan

From: Fatai Zhang [mailto:[e-mail]]
Sent: Tuesday, October 26, 2010 5:11 AM
To: Rajan Rao; Daniele Ceccarelli
Cc: Ong, Lyndon; [e-mail]
Subject: Re: [CCAMP] OSPF-TE extensions for G.709 OTN draft submitted: draft-ashok-ccamp-gmpls-ospf-g709-01

Hi Rajan,

You said there are three differences between two drafts. Now, we are coming much closer that there "may" be only one difference (i.e., backward compatibility) after Daniele gave some clarifications.

Your draft introduces a new sub-TLV SCSI and [draft-ceccarelli] introduces Generalized-ISCD sub-TLV, so I think they have the same backward compatibility, although it seems like that your draft "could" support backward compatibility.

Moreover, there are some disadvantages to use the existing ISCD +SCSI. Firstly,  as Daniele already pointed it out, it is a big issue to advertise that Max LSP bandwidth is 0 in the case of non-ODUflex supported for Node B/C, however, for Node A(G.709 V1), it must always advertise a meaningful value for the Max LSP bandwidth.
{Rajan] {
Not clear what the issue is.  We have listed 2 approaches in the slides sent out yesterday. To reiterate the case where ODUflex is not supported:
 With  backward compatibility option, the ODUflex goes in to a separate sub-TLV within SCCI. If ODUflex is not supported on a link, this sub-TLV will not be included; i.e, ISCD::Max-LSP-BW is not used for ODUflex & doesn't become zero.

Please clarify if you have a different case in mind.
}

Secondly, You have changed the meaning of "Minimum LSP Bandwidth", which will make the path computation messed. Assume OTU2 interface that supports ODU2 switching only (an example in your draft), in this case, according to the regular routing process, it will make the routing path computation function to understand that it can support ODU0 LSP.
[Rajan] {
No, we have not changed the meaning of "Minimum LSP BW".   Please elaborate on what is messed.
Following RFC 4202/4201 meaning, the example you have listed works as follows:

MinLSP-BW = ODU1  ( G.709-v1 supports 2.5G)
MaxLSP-BW = ODU2  ( Min(unreserved-BW, configured Max LSP size)  as per RFC 4202)

If you are computing path for an ODU0 connection, the rule would be

Odu0-LSP-BW >= MinLSP-BW && <= MaxLSP-BW

The computation fails for the above.   Where is it messed? Please provide a concrete example.
}

Thirdly, it can not support priority flexibility, so it is not efficient from the routing scalability perspective.
[Rajan] {
The whole idea is to stay with the GMPLS architecture already in place.  The technology specific extensions should go into SCSI. This has been done successfully for SONET/SDH case (Ref to examples in RFC 4202). Do not see a reason to deviation from the original GMPLS model.
}

Fourth, it can not differentiate termination and swiching capability for your darft.
 [Rajan] {
We do not see a need to advertise this info.  We need to understand the use cases requiring this.  Are T/S bits addressing HW limitations?
}

I am trying to recall more reasons why we need to use a new sub-TLV rather than the existing TLV, because we had lots of discussion before IETF78th meeting through mailing list and F2F, but it is a little difficult for me to dig more things from my memory at this moment, :-)~~~~
 [Rajan] {
Please list the justifications for changing original GMPLS model.
}
Best Regards

Fatai

----- Original Message -----
From: Rajan Rao<mailto:[e-mail]>
To: Daniele Ceccarelli<mailto:[e-mail]>
Cc: Ong, Lyndon<mailto:[e-mail]> ; [e-mail]<mailto:[e-mail]>
Sent: Tuesday, October 26, 2010 7:13 AM
Subject: Re: [CCAMP] OSPF-TE extensions for G.709 OTN draft submitted: draft-ashok-ccamp-gmpls-ospf-g709-01


Daniele,



Please see response inline, within { }.



Thanks

Rajan



-----Original Message-----

From: Daniele Ceccarelli [mailto:[e-mail]]

Sent: Friday, October 22, 2010 6:49 AM

To: Rajan Rao

Cc: Ong, Lyndon; [e-mail]<mailto:[e-mail]>

Subject: RE: [CCAMP] OSPF-TE extensions for G.709 OTN draft submitted: draft-ashok-ccamp-gmpls-ospf-g709-01



Hi Rajan,



Just a couple of comments on backward compatibility and bandwidth advertisement.



- Backward compatibility: no ODU advertising is definded in RFC 4202/4203, so i believe there are no backward compatibility issues.





[Rajan]  {

Agree there is no ODU advertisement defined. The need to address backwards compatibility depends on the following:



 1) are there RFC-4328 based deployments?

 2) If so, what do they use for BW advertisement?



Our fair guess is that ISCD is used for this as per RFC 4202/4203.



There are 2 ways to address the compatibility issue - either support or don't support. Based on the WG agreement we could go either way without breaking the model we have proposed. Please refer to attached slides for details.



}



Moreover in draft-ashok-ccamp-gmpls-ospf-g709 the MAX LSP bandwidth fields of the ISCD are used for unreserved bandwidth advertisement, this is a backward compatibility issue in my opinion. It states that:



      "Encoding of Max LSP Bandwidth is as follows:

       Max LSP Bandwidth = Unreserved-TS-Count x TS-Nominal-Rate"



[Rajan]  {



The unit here is in Bytes/sec. The description above (& in our draft)

-    shows that nominal rate is dependent on specific ODU layer in question.

-    shows how Max LSP bandwidth number is derived at the advertising node.

}



Moreover it states:



      "If the interface does not support ODU-flex service,

      this value should be coded as zero".



But advertising max LSP bandwidth = 0 means discarding such link in the path computation process.



[Rajan] {

Good question & very valid. I think you are bringing up a scenario where ODUflex is not supported but backward compatibility is required.



What we need here is to separate ODUflex advertisement from ISCD::Max-LSP-BW. This can be easily accomplished with a sub-TLV for ODUflex signal type in our model. BW in bytes/sec, of course.  Please refer to attached slides for details.



}



- Bandwidth advertisement: in the -04 version of our draft the bandwidth is advertised as follows:



    - Unreserved bandwidth for fixed container in number of containers

    - Unreserved bandwidth for oduflex in Bytes/sec

    - Max LSP bandwidth in Bytes/sec.



[Rajan] {

We believe advertising the number of containers is the right approach. This is exactly the proposal we have in our draft. Good to see that you have moved away from TS approach which had serious limitations.



}



Advertising the unreserved bandwidth in number of containers is the most suitable way for fixed containers (ODUk) but doesn't work with ODUflex.



[Rajan] {



 The ODUflex uses bytes/sec in our draft.  There is no disagreement here.



}

BR



The authors



-----Original Message-----

From: [e-mail] [mailto:[e-mail]] On Behalf Of Rajan Rao

Sent: venerdì 15 ottobre 2010 19.13

To: Ong, Lyndon; [e-mail]

Subject: Re: [CCAMP] OSPF-TE extensions for G.709 OTN draft submitted: draft-ashok-ccamp-gmpls-ospf-g709-01



Lyndon,



Yes, we are aware of the existing draft. There are some fundamental differences in the model we have proposed from the existing draft. The key differences are the following:



1) This draft extends existing ISCD  Vs. defining a new one

    - addresses backward compatibility issues (RFC 4202)



2) ODU containers are advertised instead of #TSs

     - simple path computation



3) Path computing node doesn't need to know per link and/or per ODU layer TS granularity

     - again, simple & straight forward path computation



Thanks

Rajan



-----Original Message-----

From: Ong, Lyndon [mailto:[e-mail]]

Sent: Thursday, October 14, 2010 7:45 AM

To: Rajan Rao; [e-mail]

Subject: RE: [CCAMP] OSPF-TE extensions for G.709 OTN draft submitted: draft-ashok-ccamp-gmpls-ospf-g709-01



Hi Rajan,



Have you looked at draft-ceccarelli-ccamp-gmpls-ospf-g709-03.txt?  It would be helpful to understand if your draft is asking for similar extensions or different functionality.



Thanks,



Lyndon







-----Original Message-----

From: [e-mail] [mailto:[e-mail]] On Behalf Of Rajan Rao

Sent: Wednesday, October 13, 2010 2:40 PM

To: [e-mail]

Subject: [CCAMP] OSPF-TE extensions for G.709 OTN draft submitted:

draft-ashok-ccamp-gmpls-ospf-g709-01



Hi,



The following draft has been submitted. Please review. Comments are appreciated.



https://datatracker.ietf.org/doc/draft-ashok-ccamp-gmpls-ospf-g709/



thanks

Authors



-----Original Message-----

From: IETF I-D Submission Tool [mailto:[e-mail]]

Sent: Wednesday, October 13, 2010 2:30 PM

To: Rajan Rao

Cc: Snigdho Bardalai; Ashok Kunjidhapatham; Khuzema Pithewan

Subject: New Version Notification for

draft-ashok-ccamp-gmpls-ospf-g709-01





A new version of I-D, draft-ashok-ccamp-gmpls-ospf-g709-01.txt has been

successfully submitted by Rajan Rao and posted to the IETF repository.



Filename:   draft-ashok-ccamp-gmpls-ospf-g709

Revision:   01

Title:            OSPF TE Extensions for Generalized MPLS (GMPLS) Control

of G.709 Optical Transport Networks

Creation_date:    2010-10-12

WG ID:            Independent Submission

Number_of_pages: 20



Abstract:

As OTN network capabilities continue to evolve, there is an increased

need to support GMPLS control for the same. [RFC4328] introduced

GMPLS signaling extensions for supporting the early version of G.709

[G.709-v1]. The basic routing considerations from signaling

perspective is also specified in [RFC4328].



The recent revision of ITU-T Recommendation G.709 [G.709-v3] and

[GSUP.43] have introduced new ODU containers (both fixed and

flexible) and additional ODU multiplexing capabilities, enabling

support for optimal service aggregation.



This document describes OSPF protocol extensions to support

Generalized MPLS (GMPLS) control for routing services over the

standardized OTU/ODU containers in support of ODU based TDM

switching. Routing support for Optical Channel Layer switching

(Lambda switching) is not covered in this document.









The IETF Secretariat.





_______________________________________________

CCAMP mailing list

[e-mail]

https://www.ietf.org/mailman/listinfo/ccamp



_______________________________________________

CCAMP mailing list

[e-mail]

https://www.ietf.org/mailman/listinfo/ccamp

________________________________
_______________________________________________
CCAMP mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/ccamp