I don't know if the pun was intended, but I do like Kireeti's comment about "bring them to light". Undoubtedly, this would be within the ITU-T (wavelength) grid! ;-)
There is, I think, some commonality in the comments and the reply in that the intent is to generalize the extensions needed for routing to accomodate non-PSC resources. I agree with the first 4 points that Jonathan made in that I think layer information must be included in routing so that important functions can be performed. I believe that the use of one or two bandwidth values was motivated by the desire to use a "lowest common denominator" attribute to generalize on path computation and avoid extensive details of links. Unfortunately, this can obscure variable adaptation on a link and the ability to determine a path at a particular adaptation (e.g., VC-3).
In variable adaptation, multiple layers can be supported on a link (actually a G.805 trail) as mentioned in the 2nd point. When a label (channel) is used at a particular layer, it has the effect of modifying the available labels in other layers supported on the same link. It is important to thus be able to reflect the layer relationship in routing.
For the issue of path computation at a particular layer, while bandwidth can give a superset of possible paths, what is needed in many contexts is a path at a specific adaptation. Because different combinations of adaptations can provide the same b/w, the result is that it infeasible paths can be computed because there is no distinction between what type of adaptation can be supported. There are times when for example, you want to have a VC-3 connection end-to-end and not just an LSP with its equivalent bandwidth because the service that was sold was specifically a VC-3.
-----Original Message-----
From: Kireeti Kompella [mailto:kireeti@juniper.net]
Sent: Saturday, March 01, 2003 14:29
To: Jonathan Sadler
Cc: iesg@ietf.org; ietf@ietf.org; ccamp@ops.ietf.org
Subject: Re: Last Call: Routing Extensions in Support of Generalized MPLS vto Proposed Standard
Hi Jonathan,
On Sun, 23 Feb 2003, Jonathan Sadler wrote:
> Please consider the following comments on these drafts:
> draft-ietf-ccamp-gmpls-routing-05.txt
> draft-ietf-ccamp-ospf-gmpls-extensions-09.txt
> Many of the comments are based on implementation experience. These
> comments are marked with a (*).
Thanks for your comments. Most of these would have more appropriate to raise in the WG Last Call, but it is certainly better to bring them to light now than never.
> ==========
>
> 1. In section 4.4.2 of draft-ietf-ccamp-gmpls-routing-05.txt, the
> operations for Packet Switch Capable (PSC) are defined. Reference is
> made to Minimum LSP bandwidth for SDH encoding. None of the examples
> in section 5 of draft-ietf-ccamp-gmpls-routing-05.txt show how this
> field should filled.
Note that all possible permutations cannot possibly be addressed; however, we will consider adding some examples here.
> 2. The mechanism for showing relationships between server and client
> layers is not generalized*. Specifically:
> - Relationships between SONET/SDH layers (ISC type: TDM) are
> implicitly stated based on the Min and Max LSP bandwidth
> advertised*. To quote draft-ietf-ccamp-gmpls-routing-05.txt:
>
> "On an interface having Standard SDH multiplexing, an LSP
> at priority p could reserve any bandwidth allowed by the
> branch of the SDH hierarchy, with the leaf and the root
> of the branch being defined by the Minimum LSP Bandwidth
> and the Maximum LSP Bandwidth at priority p."
>
> This requires node doing the route calculation to understand
> the G.707 multiplexing hierarchy. Since LSP routing is source
> routed, it could be the node doing the route calculation
> doesn't understand G.707.
This document does *not* cover how path computation is done. Of course, the information is less than useful if it cannot be used effectively. However, it has been pointed out implicitly and explicitly in several documents that the head end does *not* have to do the entire route computation -- this is one reason why there are loose hops in the ERO. See also the various drafts on multi-area TE computation.
> - Relationships between PSC-n (for IP switching) and SONET/SDH are
> explicitly specified on the encoding type specified in the PSC-n
> announcement*. However, SONET/SDH is not a single layer. It
> must be possible to identify if the PSC-n layer can be carried
> by a VC-11 trail, a VC-12 trail, a VC-3 trail, a VC-4 trail,
> or a VC-4-nc trail.
>
> Section 4.4.2 of draft-ietf-ccamp-gmpls-routing-05.txt tries
> to deal with this in the following paragraph:
>
> "On a PSC interface that supports Standard SDH encoding, an
> LSP at priority p could reserve any bandwidth allowed by
> the branch of the SDH hierarchy, with the leaf and the root
> of the branch being defined by the Minimum LSP Bandwidth
> and the Maximum LSP Bandwidth at priority p."
>
> The problem is this contradicts the following paragraph:
>
> "The Maximum LSP Bandwidth takes the place of the Maximum
> Bandwidth ([ISIS-TE], [OSPF-TE]). However, while Maximum
> Bandwidth is a single fixed value (usually simply the link
> capacity), Maximum LSP Bandwidth is carried per priority,
> and may vary as LSPs are set up and torn down."
> Specifically, how does a completely available OC-48 interface
> with VC-11 over VC-3, VC-3, and VC-4 get encoded? Based on
> the first paragraph, the encoding would be MinLSPBW=1.5M, and
> MaxLSPBW[p]=155M. Ignoring the issue that VC-11 over VC-4
> ends up being shown as supported, the second paragraph states
> that the MaxLSPBW[p] should be 2.5G.
>
> Knowing the OC-48 is completely available is useful information,
> as multiple VC-4s could be used in an LCAS group to support
> connections that need data rates over 155M*.
It is not the intent of this document to support all possible info about a link, but to support a useful *and* scalable subset. We will be guided by implementation and deployment experience to add selectively to the information. Note that the Max Reservable bandwidth field remains, and in the case that the link is not a bundle, will be set to ~2.5G.
> - The mechanism does not support describing common muli-layer
> relationships such as IP over DS1 over VC-11 or IP over DS3
> over VC-3*.
It is also not the intent of this document to provide a full description of routing info for SDH. This is a *general* document. The intent is to provide a code point for SDH to be expanded by another document. This was the model used for signaling as well: a *general* func spec, a
*general* doc for each protocol, and *SDH-specific* docs. There is an SDH specific routing doc; detailed comments are better directed there.
> - Sometimes layer relationships are described in an "inverted"
> manner*. Section 5.1 of draft-ietf-ccamp-gmpls-routing-05.txt
> states:
> "STM-16 POS Interface on a LSR
>
> Interface Switching Capability Descriptor:
> Interface Switching Capability = PSC-1
> Encoding = SDH
> Max LSP Bandwidth[p] = 2.5 Gbps, for all p"
>
> Where PSC-1 is the client of an SDH (sic) server.
>
> Section 5.5 states:
> "Interface of an opaque OXC (SDH framed) with support for
> one lambda per port/interface"
>
> " Interface Switching Capability Descriptor:
> Interface Switching Capability = LSC
> Encoding = SDH"
>
> In this case, SDH is a client of a wavelength server (LSC).
> However, unlike in section 5.1, the layer relationship is
> inverted.
Is this pointed out as a curiosity, or is there a question that needs to be addressed?
> 3. Layer specific attributes are not supported*. Specifically:
Good points. Please raise this with the SDH routing doc.
> 4. The "TDM" Interface Switching Capability presumably includes
> layers other than SONET/SDH, such as PDH* (DS1, DS3, E1, E3) and
> G.709. The interaction with these layers needs to be defined.
Ditto.
> 5. In many cases, 8 levels of priority are not necessary*. A more
> compact encoding that has a bitfield stating the priority levels
> being announced would reduce the size of the announcement.
This has been discussed elsewhere. This is the model in the base TE document; it has proven reasonable in practice. If deployment proves otherwise, this is easy to fix. For now, though, I would leave it as is.
Thanks again for your comments,
Kireeti.