Internet DRAFT - draft-ietf-ccamp-gmpls-rsvp-te-ason
draft-ietf-ccamp-gmpls-rsvp-te-ason
CCAMP Working Group J. Drake (Boeing)
Internet Draft D. Papadimitriou (Alcatel)
Proposed Category: Standard Track A. Farrel (Old Dog Consulting)
D. Brungard (ATT)
Z. Ali (Cisco)
A. Ayyangar (Juniper)
H. Ould-Brahim (Nortel)
D. Fedyk (Nortel)
Expiration Date: January 2006 July 2005
Generalized MPLS (GMPLS) RSVP-TE Signalling
in support of Automatically Switched Optical Network (ASON)
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt
Status of this Memo
By submitting this Internet-Draft, each author represents that any
applicable patent or other IPR claims of which he or she is aware
have been or will be disclosed, and any of which he or she becomes
aware will be disclosed, in accordance with Section 6 of BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that other
groups may also distribute working documents as Internet-Drafts.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt.
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
This Internet-Draft will expire on January 19, 2006.
Copyright Notice
Copyright (C) The Internet Society (2005).
Abstract
This document specifies how Generalized MPLS (GMPLS) RSVP-TE
signaling may be used and extended to satisfy the requirements of the
Automatically Switched Optical Network (ASON) architecture specified
by the ITU-T. The requirements are in a companion document
"Requirements for Generalized MPLS (GMPLS) Usage and Extensions for
Automatically Switched Optical Network (ASON)."
D.Papadimitriou et al. - Expires January 2006 1
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
In particular, this document details the mechanisms for setting up
Soft Permanent Connections (SPC), the necessary extensions in
delivering full and logical call/connection separation support, the
extended restart capabilities during control plane failures, extended
label usage and crankback signalling capability.
The mechanisms proposed in this document are applicable to any
environment (including multi-area) and for any type of interface:
packet, layer-2, time-division multiplexed, lambda or fiber
switching.
1. Table of Content
Abstract ......................................................... 1
1. Table of Content .............................................. 2
2. Conventions used in this document ............................. 3
3. Introduction .................................................. 3
3.1 Comparison with Previous Work ................................ 4
3.2 Applicability ................................................ 5
3.2.1 Network-Network Interface (I-NNI and E-NNI) ................ 6
3.2.2 User-Network Interface (UNI) ............................... 8
4. Fulfilling ASON Requirements for GMPLS Signaling .............. 8
4.1 Soft Permanent Connection (SPC) .............................. 8
4.2 Call/Connection Separation .................................. 8
4.3 Call Segments ................................................ 9
4.4 Control Plane Restart Capabilities ........................... 9
4.5 Extended Label Association ................................... 9
4.6 Crankback Signaling .......................................... 9
4.7 Additional Error Codes ....................................... 9
5. Concepts and Terms ........................................... 10
5.1 What is a Call? ............................................. 10
5.2 Hierarchy of Calls, Connections, Tunnels and LSPs ........... 10
5.3 Exchanging Access Link Capabilities ......................... 10
5.3.1 Network-initiated Calls ................................... 11
5.3.2 User-initiated Calls ...................................... 11
5.3.3 Utilizing Call Setup ...................................... 11
6. Protocol Extensions for Calls and Connections ................ 11
6.1 Call Identification ......................................... 11
6.1.1 Long Form Call Identification ............................. 12
6.1.2 Short Form Call Identification ............................ 12
6.1.3 Short Call ID Encoding .................................... 13
6.2 LINK_CAPABILITY Object ...................................... 14
6.3 Revised Message Format ...................................... 14
6.3.1 Notify Message ............................................ 15
6.4 ADMIN_STATUS Object ......................................... 15
7. Procedures in Support of Call and Connections ................ 16
7.1 Call/Connection Setup Procedures ............................ 16
7.2 Independent Call Setup ...................................... 17
7.2.1 Accepting Independent Call Setup .......................... 18
7.2.2 Rejecting Independent Call Setup .......................... 19
7.3 Adding a Connection to a Call ............................... 19
7.3.1 Adding a Reverse Direction Connection to a Call ........... 20
D.Papadimitriou et al. - Expires January 2006 2
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
7.4 Simultaneous Call/Connection Setup .......................... 20
7.4.1 Accepting Simultaneous Call/Connection Setup .............. 20
7.4.2 Rejecting Simultaneous Call/Connection Setup .............. 21
7.5 Call-Free Connection Setup .................................. 21
7.6 Call Collision .............................................. 21
7.7 Call/Connection Teardown .................................... 22
7.7.1 Removal of a Connection from a Call ....................... 22
7.7.2 Removal of the Last Connection from a Call ................ 23
7.7.3 Teardown of an "Empty" Call ............................... 23
7.7.4 Teardown of a Call with Existing Connections .............. 23
7.7.5 Teardown of a Call from the Egress ........................ 23
7.8 Control Plane Survivability ................................. 24
8. Applicability of Call and Connection Procedures .............. 25
8.1 Network-initiated Calls ..................................... 25
8.2 User-initiated Calls ........................................ 25
8.3 External Call Managers ...................................... 26
8.3.1 Call Segments ............................................. 26
9. Non-support of Call ID ....................................... 26
9.1 Non-support by External Call Managers ....................... 26
9.2 Non-support by Transit Nodes ................................ 27
9.3 Non-support by Egress Nodes ................................. 27
10. Security Considerations ..................................... 28
10.1 Call and Connection Security Considerations ................ 28
11. IANA Considerations ......................................... 28
12. Acknowledgements ............................................ 29
13. References .................................................. 29
13.1 Normative References ....................................... 30
13.2 Informative References ..................................... 30
14. Author's Addresses .......................................... 31
Appendix 1. Analysis of G.7713.2 against GMPLS RSVP-TE Signaling
Requirements in Support of ASON.................................. 32
2. Conventions used in this document
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in [RFC2119].
In addition, the reader is assumed to be familiar with the
terminology used in [RFC3471], [RFC3473], [RFC3477] and [RFC3945].
3. Introduction
This document describes how GMPLS RSVP-TE signaling [RFC3473] can be
used and extended in support of Automatically Optical Switched
Networks (ASON) as specified in the ITU-T G.8080 recommendation
[G.8080]. Note, however, that the mechanisms that it describes and
references have a larger scope than the one described in this
document.
[ASON-REQ] identifies the requirements to be covered by the
extensions to the GMPLS signaling protocols to support the
capabilities of an ASON network.
D.Papadimitriou et al. - Expires January 2006 3
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
The following are expected from the GMPLS protocol suite to realize
the needed ASON functionality:
a) support for soft permanent connection functionality
b) support for call and connection separation
c) support for call segments
d) support for extended restart capabilities during control plane
failures
e) support for extended label association
f) support for crankback capability.
This document is aligned with the [RSVP-CHANGE] process, which
requires evaluation of existing protocol functionality for achieving
the requested functionality and justification for any requested
changes or new extensions. In this context, the following summarizes
the evaluation made:
1. Signaling across the internal network-network interface (I-NNI)
and user-network interface (UNI) can be done as described in
[RFC3473] and [GMPLS-OVERLAY] respectively. Thus, the processing
of standard objects and functions (such as EXPLICIT_ROUTE Object
and RECORD_ROUTE Object) are as described in those documents.
2. The second is that any GMPLS RSVP-TE object, message or procedure
not defined in this document or in a directly referenced document
is handled exactly as described in [RFC3473], [RFC3209],
[RFC2961], and [RFC2205]. An important consideration is that the
procedures introduced by this document do not introduce any
forward or backward compatibility issue.
3. The mechanisms proposed in this document are not restricted to
LSC or TDM capable interfaces, but are equally applicable to any
packet (PSC) or layer-2 interfaces (L2SC). As a consequence, the
present document proposes ubiquitously applicable RSVP
extensions.
3.1 Comparison with Previous Work
Informational RFC [RFC3474] documents the code points for the
signaling extensions defined in [G.7713.2] to meet the requirements
of ASON Distributed Call and Connection Management (DCM) as specified
in [G.7713].
While [G.7713.2] make use of GMPLS RSVP-TE signaling, there are key
differences from the problem statement in [ASON-REQ] and the solution
it provides. These differences result from the development of a
fuller and clearer set of requirements in [G.8080] after the time
that [G.7713.2] was published and [ASON-REQ] considerations for
compatibility aspects with GMPLS [RFC3473]. These differences are
enumerated below and detailed in Appendix 1.
1. As described in [G.8080], there are various models and multiple
methods of achieving connections across multiple domains.
D.Papadimitriou et al. - Expires January 2006 4
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
[G.7713.2] is similar to a cooperative connection model between
domains, that is, there is no overall coordination, and it uses
point-to-point external NNI (E-NNI) signaling between inter-
domain border controllers (i.e. single-hop LSP). Additionally, it
requires address resolution at both border controllers regardless
of the address space used. Recent enhancements to [G.8080]
include end-to-end network capabilities based on flexible (end-
to-end) path selection to support optimal route selection i.e.
source-based re-routing and crankback.
To provide for these enhancements and future capabilities (e.g.,
VPNs), [ASON-REQ] is based on an inter-domain model using an end-
to-end call model, modeling multiple domains as one virtual
network, and optional one-time (ingress) address resolution
(optional, if multiple address families are needed). Note that
this model is the same model used by [RFC3471], [RFC3473] and
[GMPLS-OVERLAY].
2. [G.7713.2] distinguishes between use of [G.7713.2] for ASON
networks and use of [RFC3473] for GMPLS networks; however, no
compatibility aspects are addressed. [ASON-REQ] addresses ASON
requirements for GMPLS networks. Backward compatibility allows
for the coexistence of nodes supporting GMPLS RSVP-TE [RFC3473]
with nodes supporting GMPLS RSVP-TE for ASON (as described in
this document).
[ASON-REQ] requires that for any new and existing GMPLS features,
[RFC3473] transit nodes do not need to be updated and do not need
to modify their behavior to support the end-to-end features of
ASON. The solution provided by [G.7713.2] is not backward
compatible with [RFC3473]. Moreover, [G.7713.2] can not be used
in a network with [RFC3473], as incorrect network behavior will
result.
3. While existing GMPLS signalling [RFC3473] supports Soft Permanent
Connections (SPCs), [G.7713.2] defines a new mechanism to support
SPCs, and this new mechanism is incompatible with [RFC3473].
4. [G.7713.2] does not support full call and connection separation,
multiple connections per call, or ingress/egress node capability
negotiation prior to connection establishment.
5. [G.7713.2] does not support call segment signaling mechanisms, as
required in [G.8080] and [G.7713].
6. [G.7713.2] defines control plane restart capabilities that are
incompatible with those described in [RFC3473].
7. [G.7713.2] does not support crankback signaling mechanisms
[GMPLS-CRANK], as required in [G.8080] and [G.7713].
3.2 Applicability
D.Papadimitriou et al. - Expires January 2006 5
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
The requirements placed on the signaling plane of an optical network
to support the capabilities of an Automatically Switched Optical
Network (see [ASON-REQ]) apply at both the network-network interface
(NNI) and the user-network interface (UNI).
Some extensions to the core signaling features (see [RFC3473]) are
required in support of some of the ASON requirements. [GMPLS-OVERLAY]
defines a common set of standard procedures at the user-network
interface (UNI). Other documents referenced in specific subsections
of this document define specific protocol extensions in support of
specific ASON requirements.
3.2.1 Network-Network Interface (I-NNI and E-NNI)
At the NNI, the ingress and egress core nodes play a full part in the
GMPLS network from a signaling point of view. Applicability of GMPLS
RSVP-TE signaling at the I-NNI is implicitly detailed in [RFC3471]
and [RFC3473]. Routing information is fully or partially distributed
through this multi-vendor interface.
The following paragraphs further detail the applicability of
[RFC3471] and [RFC3473] mechanisms at the E-NNI. Note also that the
use of these RFCs at the E-NNI does not preclude the use of another
signaling protocol for the I-NNI as long as an inter-working function
is provided by the non-GMPLS domain. Routing information may be fully
or partially distributed through this interface.
The basic GMPLS RSVP-TE operations at the E-NNI reference point
involves (as inspired from [GMPLS-OVERLAY]):
1. Addressing
Two adjacent egress/ingress core nodes must share the same address
space, which is used by GMPLS E-NNI signaling. A set of egress/
ingress core node tuples share the same address space if the ingress
or ingress core node in the set could exchange GMPLS RSVP-TE messages
among themselves. Within a control domain, the address space used by
the core nodes to communicate among themselves MAY, but need not be
shared with the address space used by any of the egress/ingress core
node tuples.
A core node is identified by either a single IP address representing
its Node ID, or by one or more un/numbered TE links that interconnect
core-nodes. A core node needs only to know (and track) the interface
addresses and/or Node IDs of client nodes to which incoming messages
are directed.
Links may be either numbered or unnumbered. Further, links may be
bundled or unbundled. See [BUNDLE] and [RFC3477], respectively.
(IF_ID) RSVP_HOP object processing at E-NNI boundaries follows the
rules defined in [RFC3473].
2. ERO Processing
D.Papadimitriou et al. - Expires January 2006 6
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
An ingress core node MAY receive and potentially reject a Path
message that contains an ERO. Such behavior is controlled by
(hopefully consistent) configuration. If an ingress core node rejects
a Path message due to the presence of an ERO it SHOULD return a
PathErr message with an error code of "Unknown object class" toward
the sender. This causes the path setup to fail.
Further an ingress core node MAY accept EROs that include a sequence
of [<egress core node, ingress core node>]. This is to support
explicit label control on the egress core node interface. Incoming
EROs may also include a combination of the latter with sequence of
loose ingress core node addresses and/or AS numbers. If an ingress
core node rejects a Path message due to the presence of an ERO not of
the permitted format it SHOULD return a PathErr message with an error
code of Bad Explicit Route Object as defined in [RFC3209].
- Path Message without ERO: when an ingress core node receives a Path
message from an egress core node that contains no ERO, it MUST
calculate a route to the destination and include that route in a
ERO, before forwarding the PATH message. One exception would be if
the egress core node were also adjacent to this core node. If no
route can be found, the ingress core node SHOULD return a PathErr
message with an Error code and value of 24,5 - "No route available
toward destination".
- Path Message with ERO: when an ingress core node receives a Path
message from an egress core node that contains an ERO, the ingress
core node SHOULD verify the route against its topology database
before forwarding the PATH message. If the route is not viable,
then a PathErr message with an error code and value of 24,5 - "No
route available toward destination" should be returned.
3. RRO Processing
An egress or an ingress core node MAY include an RRO and MAY remove
the RRO from the received Path message before forwarding it. Further
an egress or an ingress core node MAY remove the RRO from a Resv
message before forwarding it. Such behavior is controlled by
(hopefully consistent) configuration.
Further an ingress core node MAY edit the RRO in a Resv message such
that it includes only the subobjects from the egress core node
through the ingress core node of a neighboring E-NNI. This is to
allow the ingress core node to be aware of the selected link and
labels on the far end of the connection traversing this network.
4. Notification
An ingress core node MAY include a NOTIFY_REQUEST object in both the
Path and Resv messages it forwards. An ingress node MAY remove the
NOTIFY_REQUEST object from the Path and Resv message before
forwarding it. An egress node MAY remove the NOTIFY_REQUEST object
from the Path and Resv message before forwarding it. Core nodes may
D.Papadimitriou et al. - Expires January 2006 7
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
send Notification messages to ingress/egress core nodes, which have
included the NOTIFY_REQUEST object.
Note: the use of the Notify message for independent Call setup as
defined in this document extends the one specified in [RFC-3473].
3.2.2 User-Network Interface (UNI)
At the UNI, the ingress and/or the egress nodes are not full players
in the GMPLS network. Signaling information may be filtered and
substituted by the network. This process is described in [GMPLS-
OVERLAY]. Routing information leaked to the ingress/egress nodes is
very limited.
The ingress node may initiate an LSP setup/teardown request to the
network using standard GMPLS procedures. The modifications to
behavior described in [GMPLS-OVERLAY] apply to the nodes within the
network (in particular, the network edge nodes) and not ingress or
egress nodes.
4. Fulfilling ASON Requirements for GMPLS Signaling
This section briefly describes how the requirements identified in
[ASON-REQ] are met by existing GMPLS specifications, by procedures
described in this document, or by other means.
4.1 Soft Permanent Connection (SPC)
A Soft Permanent Connection (SPC) is defined as a combination of a
permanent connection at the network edges and a switched connection
within the network.
SPC setup is provided using Explicit Label Control as specified in
[RFC3473], with the augmented procedures described in [GMPLS-
OVERLAY].
4.2 Call/Connection Separation
The call concept for optical networks is defined in [G.8080] where it
is used to deliver the following capabilities:
- Verification and identification of the call initiator (prior to
LSP setup)
- Support of virtual concatenation with diverse path component LSPs
- Multiple LSP association with a single call (note aspects related
to recovery are detailed in [GMPLS-FUNCT] and [GMPLS-E2E])
- Facilitate control plane operations by allowing operational status
change of the associated LSP.
D.Papadimitriou et al. - Expires January 2006 8
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
Procedures and protocol extensions to support Call setup, and the
association of Calls with Connections are described in sections 5 and
onwards of this document.
4.3 Call Segments
Call segments capabilities MUST be supported by both independent call
setup and simultaneous call/connection setup.
Procedures and (GMPLS) RSVP-TE signaling protocol extensions to
support call segments are described in sections 8.4.1 of this
document.
4.4 Control Plane Restart Capabilities
Restart capabilities are provided by GMPLS RSVP-TE signaling in case
of control plane failure including nodal and control channel faults.
The handling of node and control channels faults is described in
[RFC3473] Section 9. No additional RSVP mechanisms or objects are
required to fulfill the ASON control plane restart capabilities.
However, it should be noted that restart considerations must form
part of each of the procedures referenced from or described in this
document.
4.5 Extended Label Association
Dynamic discovery of label associations as described in [ASON-REQ]
can be either performed through manual provisioning or using the Link
Management Protocol [LMP] capabilities.
4.6 Crankback Signaling
Crankback signaling allows a connection setup request to be retried
on an alternate path that detours around a blocked link or node upon
a setup failure, for instance, because a link or a node along the
selected path has insufficient resources. Crankback mechanisms may
also be applied during connection recovery by indicating the location
of the failed link or node. This would significantly improve the
successful recovery ratio for failed connections, especially in
situations where a large number of setup requests are simultaneously
triggered.
Crankback mechanisms for (GMPLS) RSVP-TE signaling are covered in a
dedicated companion document [GMPLS-CRANK]. That document is intended
to fulfill all the corresponding ASON requirements as well as
satisfying any other crankback needs.
4.7 Additional Error Codes
Error codes corresponding to the mechanisms defined in this document
are specified along each section and summarized in Section 11.
D.Papadimitriou et al. - Expires January 2006 9
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
5. Concepts and Terms
The concept of a Call and a Connection are discussed in the ASON
architecture [G.8080]. This section is not intended as a substitute
for that document, but is a brief summary of the key terms and
concepts.
5.1 What is a Call?
A Call is an agreement between endpoints possibly in cooperation with
the nodes that provide access to the network. Call setup may include
capability exchange, policy, authorization and security.
A Call is used to facilitate and manage a set of Connections that
provide end to end data services. While Connections require state to
be maintained at nodes along the data path within the network, Calls
do not involve the participation of transit nodes except to forward
the Call management requests as transparent messages.
A Call may be established and maintained independently of the
Connections that it supports.
5.2 A Hierarchy of Calls, Connections, Tunnels and LSPs
Clearly there is a hierarchical relationship between Calls and
Connections. One or more Connections may be associated to a Call. A
Connection may not be part of more than one call. A Connection may,
however, exist without a Call.
In GMPLS RSVP-TE [RFC3473], a Connection is identified with a GMPLS
TE Tunnel. Commonly a Tunnel is identified with a single LSP, but it
should be noted that for protection, load balancing and many other
functions, a Tunnel may be supported by multiple parallel LSPs. The
following identification reproduces this hierarchy:
- Call IDs are unique within the context of the pair of addresses
that are the source and destination of the Call.
- Tunnel IDs are unique within the context of the Session (that is
the destination of the Tunnel). Applications may also find it
convenient to keep the Tunnel ID unique within the context of a
Call.
- LSP IDs are unique within the context of a Tunnel.
Note that the Call_ID value of zero is reserved and MUST NOT be used
during LSP-independent call establishment.
Throughout the remainder of this document, the terms LSP and Tunnel
are used interchangeably with the term Connection. The case of a
Tunnel that is supported by more than one LSP is covered implicitly.
D.Papadimitriou et al. - Expires January 2006 10
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
5.3 Exchanging Access Link Capabilities
It is useful for the ingress node of an LSP to know the link
capabilities of the link between the network and the egress node.
This information may allow the ingress node to tailor its LSP request
to fit those capabilities and to better utilize network resources
with regard to those capabilities.
In particular, this may be used to achieve end-to-end spectral
routing attribute negotiation for signal quality negotiation (such as
BER) in photonic environments where network edges are signal
regeneration capable. Similarly, it may be used to provide end-to-end
spatial routing attribute negotiation in multi-area routing
environments, in particular, when TE links have been bundled based on
technology specific attributes.
Call setup may provide a suitable mechanism to exchange information
for this purpose, although several other possibilities exist.
5.3.1 Network-initiated Calls
In this case, there may be no need to distribute additional link
capability information over and above the information distributed by
the TE and GMPLS extensions to the IGP. Further, it is possible that
future extensions to these IGPs will allow the distribution of more
detailed information including optical impairments.
5.3.2 User-initiated Calls
In this case, edge link information may not be visible within the
core network, nor (and specifically) at other edge nodes. This may
prevent an ingress from requesting suitable LSP characteristics to
ensure successful LSP setup.
Various solutions to this problem exist including the definition of
static TE links (that is, not advertised by a routing protocol)
between the core network and the edge nodes. Nevertheless, special
procedures may be necessary to advertise edge TE link information to
the edge nodes outside of the network without advertising the
information specific to the contents of the network.
In the future, when the requirements are understood on the
information that needs to be supported, TE extensions to EGPs may be
defined that provide this function.
5.3.3 Utilizing Call Setup
When IGP and EGP solutions are not available at the UNI, there is
still a requirement to have, at the local edge nodes, the knowledge
of the remote edge link capabilities.
The Call setup procedure provides an opportunity to discover edge
link capabilities of remote edge nodes before LSP setup is attempted.
D.Papadimitriou et al. - Expires January 2006 11
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
The LINK CAPABILITY object is defined to allow this information to be
exchanged. The information that is included in this object is similar
to that distributed by GMPLS-capable IGPs (see [GMPLS-RTG]).
6. Protocol Extensions for Calls and Connections
This section describes the protocol extensions needed in support of
Call identification and management of Calls and Connections.
Procedures for the use of these protocol extensions are described in
section 7.
6.1 Call Identification
As soon as the concept of a call is introduced, it is necessary to
support some means of identifying the call. This becomes particularly
important when calls and connections are separated and connections
must contain some reference to the call.
According to [ASON-REQ], a call may be identified by a sequence of
bytes that may have considerable (but not arbitrary) length. A Call
ID of 40 bytes would not be unreasonable. It is not the place of this
document to supply rules for encoding or parsing Call IDs, but it
must provide a suitable means to communicate Call IDs within the
protocol. The full call identification as required by ASON is
referred to as the long Call ID.
The Call_ID is only relevant at the sender and receiver nodes.
Maintenance of this information in the signaling state is not
mandated at any intermediate node. Thus no change in [RFC3473]
transit implementations is required and there are no backward
compatibility issues. Forward compatibility is maintained by using
the existing default values to indicate that no Call processing is
required.
6.1.1 Long Form Call Identification
The "Session Name" attribute of the SESSION_ATTRIBUTE Object is used
to carry the long form of the Call ID.
A unique value per call is inserted in the "Session Name" field by
the initiator of the call. Subsequent network nodes MAY inspect this
object and MUST forward this object transparently across network
interfaces until reaching the egress node. Note that the structure of
this field MAY be the object of further formatting depending on the
naming convention(s). However, [RFC3209] defines the "Session Name"
field as a Null padded display string, and that any formatting
conventions for the Call ID must be limited to this scope.
6.1.2 Short Form Call Identification
The connections (LSPs) associated with a call need to carry a
reference to the call - the Call ID. Each LSP MAY carry the full long
Call ID in the "Session Name" of the SESSION ATTRIBUTE object to
D.Papadimitriou et al. - Expires January 2006 12
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
achieve this purpose. However, existing (and future) implementations
may need to place other strings in this field (in particular, the
field is currently intended to provide the Session Name). To allow
for this possibility a new field is added to the signaling protocol
to identify an individual LSP with the Call to which it belongs.
The new field is a 16-bit identifier (unique within the context of
the address pairing provided by the Tunnel_End_Point_Address and the
Sender_Address) that MUST be exchanged during Call initialization and
is used on all subsequent LSP setups that are associated with the
Call. This identifier is known as the short Call ID and is encoded as
described in Section 6.1.3. When relevant, the Call Id MUST NOT be
used as part of the processing to determine the session to which an
RSVP signaling message applies. This does not generate any backward
compatibility issue since the reserved field of the SESSION object
defined in [RFC3209] MUST NOT be examined on receipt.
In the unlikely case of short Call_ID exhaustion, local node policy
decides upon specific actions to be taken. Local policy details are
outside of the scope of this document.
6.1.3 Short Form Call ID Encoding
The short Call ID is carried in a 16-bit field in the SESSION object
used during Call and LSP setup. The field used was previously
reserved (MUST be set to zero on transmission and ignored on
receipt). This ensures backward compatibility with nodes that do not
utilize calls.
The figure below shows the new version of the object.
Class = SESSION, Class-Num = 1, C-Type = 7(IPv4)/8(IPv6)
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| IPv4/IPv6 Tunnel end point address |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Call_ID | Tunnel ID |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Extended Tunnel ID |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
IPv4/IPv6 Tunnel End Point Address: 32 bits/128 bits (see [RFC3209])
Call_ID: 16 bits
A 16-bit identifier used in the SESSION object that remains
constant over the life of the call. The Call_ID value MUST be
set to zero when there is no corresponding call.
Tunnel ID: 16 bits (see [RFC3209])
D.Papadimitriou et al. - Expires January 2006 13
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
Extended Tunnel ID: 32 bits/128 bits (see [RFC3209])
6.2 LINK_CAPABILITY object
The LINK CAPABILITY object is introduced to support link capability
exchange during Call setup. This optional object includes the bundled
link local capabilities of the call initiating node (or terminating
node) indicated by the source address of the Notify message.
The Class Number is selected so that the nodes that do not recognize
this object drop it silently. That is, the top bit is set and the
next bit is clear.
This object has the following format:
Class-Num = TBA (form 10bbbbbb), C_Type = 1
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
// (Subobjects) //
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
The contents of the LINK_CAPABILITY object is defined as series of
variable-length data items called subobjects. The subobject format is
defined in [RFC3209].
The following subobjects are currently defined:
- Type 1: the link local IPv4 address (numbered bundle) using the
format defined in [RFC3209]
- Type 2: the link local IPv6 address (numbered bundle) using the
format defined in [RFC3209]
- Type 4: the link local identifier (unnumbered links and bundles)
using the format defined in [RFC3477]
- Type 64: the Maximum Reservable Bandwidth corresponding to this
bundle (see [BUNDLE])
- Type 65: the interface switching capability descriptor (see
[GMPLS-RTG]) corresponding to this bundle (see also [BUNDLE]).
Note: future revisions of this document may extend the above list.
This object MAY also be used to exchange more than one bundled link
capability. In this case, the following ordering MUST be followed:
one identifier subobject (Type 1, 2 or 4) MUST be inserted before any
capability subobject (Type 64 or 65) to which it refers.
6.3 Revised Message Formats
The Notify message is enhanced (and referred thereby to as an
unsolicited Notify message) to support Call establishment and
D.Papadimitriou et al. - Expires January 2006 14
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
teardown of Calls that operate independent of LSPs. See section 7 for
a description of the procedures.
6.3.1 Notify Message
The Notify message is modified in support of Call establishment by
the optional addition of the LINK CAPABILTY object. Further, the
SESSION ATTRIBUTE object is added to the <notify session> sequence to
carry the long Call ID. The presence of the SESSION ATTIBUTE object
MAY be used to distinguish a Notify message used for Call management.
The <notify session list> MAY be used to setup simultaneously
multiple Calls.
The format of the Notify Message is as follows:
<Notify message> ::= <Common Header> [ <INTEGRITY> ]
[[ <MESSAGE_ID_ACK> | <MESSAGE_ID_NACK>]...]
[ <MESSAGE_ID> ]
<ERROR_SPEC>
<notify session list>
<notify session list> ::= [ <notify session list> ] <notify session>
<notify session> ::= <SESSION> [ <ADMIN_STATUS> ]
[ <POLICY_DATA>...]
[ <LINK_CAPABILITY> ]
[ <SESSION_ATTRIBUTE> ]
[ <sender descriptor> | <flow descriptor> ]
<sender descriptor> ::= see [RFC3473]
<flow descriptor> ::= see [RFC3473]
6.4 ADMIN_STATUS Object
Messages (such as Notifys, Paths, etc.) exchanged for Call control
and management purposes carry a specific new bit (the Call Management
or C bit) in the ADMIN STATUS object.
The format and the contents of the ADMIN_STATUS object are both
dictated by [RFC3473] in favor of [RFC3471]. The new "C" bit is added
as shown below.
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|R| Reserved |C|T|A|D|
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Reflect (R): 1 bit - see [RFC3471]
Testing (T): 1 bit - see [RFC3471]
Administratively down (A): 1 bit - see [RFC3471]
Deletion in progress (D): 1 bit - see [RFC3471]
D.Papadimitriou et al. - Expires January 2006 15
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
Call Management (C): 1 bit
This bit is set when the message is being used to control
and manage a Call.
The procedures for the use of the C bit are described in section 7.
Note that the use of the C bit may appear as redundant since Call
setup can be distinguished by the presence of the SESSION ATTRIBUTE
object in a Notify message or an non-zero short Call ID value in a
Path message. However, in the case of lost messages and node restart,
this further distinction is useful to distinguish Path messages that
set up Calls from Path messages that belong to calls.
7. Procedures in Support of Calls and Connections
7.1 Call/Connection Setup Procedures
This section describes the processing steps for call and connection
setup.
There are four cases considered:
- A Call and Connection may be established simultaneously. That is,
a Connection may be established and designated as belonging to a
new Call. It is an implementation decision how the work a the
ingress and egress points is split and whether the qualities of
the Call are policed before, after or at the same time as those of
the Connection. In the event that the establishment of either the
Call or the Connection fails, an error is returned as described in
Section 7.4.2 and neither is set up.
- A Call can be set up on its own. That is, without any associated
Connection. It is assumed that Connections will be added to the
Call at a later time, but this is neither a requirement nor
a constraint.
- A Connection may be added to an existing Call. This may happen if
the Call was set up without any associated Connections, or if a
further Connection is added to a Call that already has one or more
associated Connections.
- A Connection may be established without any reference to a Call.
This encompasses the previous LSP setup procedure.
Note that a Call MAY NOT be imposed upon a Connection that is already
established. To do so would require changing the short Call Id in the
SESSION OBJECT of the existing LSPs and this would constitute a
change in the Session Identifier. This is not allowed by existing
protocol specifications.
D.Papadimitriou et al. - Expires January 2006 16
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
Call and Connection teardown procedures are described later in
Section 7.7.
7.2 Independent Call Setup
It is possible to set up a Call before, and independent of, LSP
setup. A Call setup without LSPs MUST follow the procedure described
in this section.
Prior to the LSP establishment, Call setup MAY necessitate
verification of the link status and link capability negotiation
between the Call ingress node and the Call egress node. The procedure
described below is applied only once for a Call and hence only once
for the set of LSPs associated with a Call.
The Notify message (see [RFC3473]) is used to signal the Call setup
request and response. The new Call Management (C) bit in the
ADMIN_STATUS object is used to indicate that this Notify is managing
a Call. The Notify message is sent with source and destination
IPv4/IPv6 address set to any of the routable ingress/egress node
addresses respectively.
At least one session MUST be listed in the <notify session list> of
the Notify message. In order to allow for long identification of the
Call the SESSION_ATTRIBUTE object is added as part of the <notify
session list>. Note that the ERROR SPEC object is not relevant in
Call setup and MUST carry the Error Code zero ("Confirmation") to
indicate that there is no error.
During Call setup, the ADMIN STATUS object is sent with the following
bits set. Bits not listed MUST be set to zero.
R - to cause the egress to respond
C - to indicate that this message is managing a Call.
The SESSION, SESSION ATTRIBUTE, SENDER_TEMPLATE, SENDER_TSPEC objects
included in the <notify session> of the Notify message are built as
follows:
- The SESSION object includes as Tunnel_End_Point_Address any of the
call terminating (egress) node's IPv4/IPv6 routable addresses. The
Call_ID is set to a non-zero value unique within the context of
the address pairing provided by the Tunnel_End_Point_Address and
the Sender_Address from the SENDER TEMPLATE object (see below).
Note that the Call_ID value of zero is reserved and MUST NOT be
used during LSP-independent call establishment. The Tunnel_ID of
the SESSION object is not relevant for this procedure and SHOULD
be set to zero. The Extended_Tunnel_ID of the SESSION object is
not relevant for this procedure and MAY be set to zero or to an
address of the ingress node.
- The SESSION ATTRIBUTE object contains priority flags. Currently no
use of these flags is envisioned, however, future work may
D.Papadimitriou et al. - Expires January 2006 17
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
identify value is assigning priorities to Calls; accordingly the
Priority fields MAY be set to non-zero values. None of the Flags
in the SESSION ATTRIBUTE object are relevant to this process and
this field SHOULD be set to zero. The Session Name field is used
to carry the long Call Id as described in Section 6.
- The SENDER_TEMPLATE object includes as Sender Address any of the
call initiating (ingress) node's IPv4/IPv6 routable addresses. The
LSP_ID is not relevant and SHOULD be set to zero.
- The bandwidth value inserted in the SENDER_TSPEC and FLOWSPEC
objects MUST be ignored upon receipt and SHOULD be set to zero
when sent.
Additionally, ingress/egress nodes that need to communicate their
respective link local capabilities may include a LINK_CAPABILITY
object in the Notify message.
The receiver of a Notify message may identify whether it is part of
Call management or reporting an error by the presence or absence of
the SESSION ATTRIUBTE object in the <notify session list>. Full
clarity, however, may be achieved by inspection of the new Call
Management (C) bit in the ADMIN STATUS object.
Note that the POLICY_DATA object may be included in the <notify
session list> and may be used to identify requestor credentials,
account numbers, limits, quotas, etc. This object is opaque to RSVP,
which simply passes it to policy control when required.
Message IDs MUST be used during independent Call setup.
7.2.1 Accepting Independent Call Setup
A node that receives a Notify message carrying the ADMIN STATUS
object with the R and C bits set is being requested to set up a Call.
The receiver may perform authorization and policy according to local
requirements.
If the Call is acceptable, the receiver responds with a Notify
message reflecting the information from the Call request with two
exceptions.
- The responder removes any LINK CAPABLITY object that was received
and MAY insert a LINK CAPABILITY object that describes its own
access link.
- The ADMIN STATUS object is sent with only the C bit set. All other
bits MUST be set to zero.
The responder MAY use the Message ID object to ensure reliable
delivery of the response. If no Message ID Acknowledgement is
received after the configured number of retries, the responder should
D.Papadimitriou et al. - Expires January 2006 18
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
continue to assume that the Call was successfully established. Call
liveliness procedures are covered in section 7.8.
7.2.2 Rejecting Independent Call Setup
Call setup may fail or be rejected.
If the Notify message can not be delivered, no Message ID
acknowledgement will be received by the sender. In the event that the
sender has retransmitted the Notify message a configurable number of
times without receiving a Message ID Acknowledgement (as described in
[RFC3473]), the initiator SHOULD declare the Call failed and SHOULD
send a Call teardown request (see section 7.7).
It is also possible that a Message ID Acknowledgement is received but
no Call response Notify message is received. In this case, the
initiator MAY re-send the Call setup request a configurable number of
times (see Section 7.8) before declaring the Call has failed. At this
point the initiator MUST send a Call teardown request (see Section
7.7).
If the Notify message cannot be parsed or is in error it MAY be
responded to with a Notify message carrying the error code 13
("Unknown object class") or 14 ("Unknown object C-Type").
The Call setup may be rejected by the receiver because of security,
authorization or policy reasons. Suitable error codes already exist
and can be used in the ERROR SPEC object included in the Notify
message sent in response.
Error response Notify messages SHOULD also use the Message ID object
to achieve reliable delivery. No action should be taken on the
failure to receive a Message ID Acknowledgement after the configured
number of retries.
7.3 Adding a Connections to a Call
Once a Call has been established, LSPs can be added to the Call.
Since the short Call ID is part of the SESSION Object, any LSP that
has the same Call ID value in the SESSION Object belongs to the same
Call. There will be no confusion between LSPs that are associated
with a Call and those which are not since the Call ID value MUST be
equal to zero for LSPs which are not associated with a Call.
LSPs for different Calls can be distinguished because the Call ID is
unique within the context of the source address (in the SENDER
TEMPLATE object) and the destination address (in the SESSION object).
Ingress and egress nodes may group together LSPs associated with the
same call and process them as a group according to implementation
requirements. Transit nodes need not be aware of the association of
multiple LSPs with the same Call.
D.Papadimitriou et al. - Expires January 2006 19
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
The ingress node MAY choose to set the "Session Name" of an LSP to
match the long Call ID of the associated Call and the "Session Name"
MAY still be used to distinguish between virtually concatenated LSPs
belonging to the same Call. Thus, there is not necessarily a one-to-
one mapping between the "Session Name" of an LSP and the short
Call_ID.
The C bit of the ADMIN STATUS object MUST NOT be set on LSP messages.
7.3.1 Adding a Reverse Direction LSP to a Call
Note that once a Call has been established it is symmetric. That is,
either end of the Call may add LSPs to the Call.
Special care is needed when managing LSPs in the reverse direction
since the addresses in the SESSION and SENDER TEMPLATE are reversed.
However, since the short Call ID is unique in the context of a given
ingress-egress address pair it may safely be used to associate the
LSP with the Call.
7.4 Simultaneous Call/Connection Setup
It is not always necessary to establish a Call before adding
Connections to the Call. Where the features made available by
independent Call setup are not required, a simplification can be made
by establish a Call at the same time as the first Connection
associated to the Call.
Simultaneous Call and LSP setup requires the usage of Call
identification and an indication that a Call is being managed. No
other protocol mechanisms beyond those described in [RFC3473] are
needed. Normal RSVP-TE GMPLS processing takes place.
The Path message used to simultaneously set up the Call and LSP MUST
carry the ADMIN STATUS object with the R and C bits set. Other bits
may be set or cleared according to the requirements of LSP setup. The
D bit MUST NOT be set.
The Path message MUST also carry the long Call ID in the Session Name
field of the SESSION ATTRIBUTE object as described above. This field
is not available to contain a Session Name distinct from the Call ID.
A non-zero short Call ID MUST be placed in the new Call ID field of
the SESSION object as described above. The reserved value of zero is
used when the LSP is being set up with no association to a Call.
7.4.1 Accepting Simultaneous Call/Connection Setup
A Path message that requests simultaneous Call and Connection setup
is subject to local authorization and policy procedures applicable to
Call establishment in addition to the standard procedures associated
with LSP setup described in [RFC3473].
D.Papadimitriou et al. - Expires January 2006 20
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
If the Call and LSP setup is to be accepted, a Resv message is
returned. The Resv message MUST carry the ADMIN STATUS object with
the R bit clear and the C bit set. Other bits may be set or cleared
according to the requirements of LSP setup. The D bit MUST NOT be
set.
The Call ID MUST be reflected in the SESSION object.
7.4.2 Rejecting Simultaneous Call/Connection Setup
The Path message that is sent to set up a Call and Connection
simultaneously may fail or be rejected.
Failures may include all those reasons described in [RFC3473].
Additionally, policy and authorization reasons specifically
associated with Call setup may cause the Path message to be rejected.
The PathErr message is issued to signal such failures and no new
error codes are required. It is RECOMMENDED that the procedures for
PathErr with state removal described in [RFC3473] is used during Call
setup failure processing.
7.5 Call-Free Connection Setup
It continues to be possible to set up LSPs as per [RFC3473] without
associating them with a Call. If the short Call ID in the SESSION
Object is set to zero, there is no associated Call and the Session
Name field in the SESSION ATTRIBUTE object SHOULD be interpreted
simply as the name of the session (see [RFC3209]).
The new C bit in the ADMIN STATUS object SHOULD be set to zero in
such messages and MUST be ignored if the Call ID is zero.
7.6 Call Collision
Since Calls are symmetrical, it is possible that both ends of a call
will attempt to establish a Call with the same long Call ID at the
same time. This is only an issue if the source and destination
address pair matches. This situation can be avoided by applying some
rules to the contents of the long Call ID, but that is outside the
scope of this document.
If a node that has sent a Call setup request and has not yet received
a response, itself receives a Call setup request with the same long
Call ID and matching source/destination addresses it should process
as follows.
- If its source address is numerically greater than the remote
source address, it MUST discard the received message and continue
to wait for a response to its setup request.
- If its source address is numerically smaller than the remote
source address, it MUST discard state associated with the Call
D.Papadimitriou et al. - Expires January 2006 21
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
setup that it initiated, and MUST respond to the received Call
setup.
In the second case, special processing is necessary if simultaneous
Call and Connection establishment was being used. Firstly, the node
that is discarding the Call that it initiated MUST send a PathTear
message to remove state from transit nodes. Secondly, this node may
want to hold onto the Connection request and establish an LSP once
the Call is in place since only the Call that it was trying to
establish has been set up by the destination - the Connection may
still be required.
A further possibility for contention arises when Call IDs are
assigned by a pair of nodes for two distinct Calls that are set up
simultaneously. In this event a node receives a Call setup request
carrying a short Call ID that matches one that it previously sent for
the same address pair. The following processing MUST be followed.
- If the receiver's source address is numerically greater than the
remote source address, the receiver returns an error (Notify
message or PathErr message as appropriate) with the new Error Code
"Call Management" (TBD) and the new Error Value "Call ID
Contention" (TBD).
- If the receiver's source address is numerically less than the
remote source address, the receiver accepts and processes the Call
request. It will receive an error message sent as described above,
and at that point it selects a new short Call ID and re-sends the
Call setup request.
Note: these procedures apply for any combination of independent and
simultaneous call establishment.
7.7 Call/Connection Teardown
As with Call/Connection setup, there are several cases to consider.
- Removal of a Connection from a Call
- Removal of the last Connection from a Call
- Teardown of an "empty" Call
The case of tearing down an LSP that is not associated with a Call
does not need to be examined as it follows exactly the procedures
described in [RFC3473].
7.7.1 Removal of a Connection from a Call
An LSP that is associated with a Call may be deleted using the
standard procedures described in [RFC3743]. No special procedures are
required.
Note that it is not possible to remove an LSP from a Call without
deleting the LSP. It is not valid to change the short Call ID from
D.Papadimitriou et al. - Expires January 2006 22
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
non-zero to zero since this involves a change to the SESSION object,
which is not allowed.
7.7.2 Removal of the Last Connection from a Call
When the last LSP associated with a Call is deleted the question
arises as to what happens to the Call. Since a Call may exist
independently of Connections, it is not always acceptable to say that
the removal of the last LSP from a Call removes the Call.
If the Call was set up using independent Call setup (that is, using a
Notify message) the removal of the last LSP does not remove the Call
and the procedures described in the next section MUST be used to
delete the Call.
If the Call was set up using simultaneous Call/Connection
establishment, the removal of the last LSP does remove the Call and
the Call ID becomes invalid.
7.7.3 Teardown of an "Empty" Call
When all LSPs have been removed from a Call that was set up
independent of Connections, the Call may be torn down or left for use
by future LSPs.
Deletion of such Calls is achieved by sending a Notify message just
as for Call setup, but the ADMIN STATUS object carries the R, D and C
bits on the teardown request and the D and C bits on the teardown
response. Other bits MUST be set to zero.
When a Notify message is sent for deleting a call and the initiator
does not receive the corresponding reflected Notify message (or
possibly even the Message ID Ack), the initiator MAY retry the
deletion request using the same retry procedures as used during Call
establishment. If no response is received after full retry, the node
deleting the Call MAY declare the Call deleted, but under such
circumstances the node SHOULD avoid re-using the long or short Call
IDs for at least the five times the Notify refresh period.
7.7.4 Teardown of a Call with Existing Connections
If a Notify request with the D bit of the ADMIN STATUS object set is
received for a Call for which LSPs still exist, the request MUST be
rejected with the Error Code "Call Management" (TBD) and Error Value
"Connection Still Exists" (TBD).
7.7.5 Teardown of a Call from the Egress
Since Calls are symmetric they may be torn down from the ingress or
egress.
If the Call was established using simultaneous Call/Connection setup
the removal of the last LSP deletes the Call. This, regardless of
D.Papadimitriou et al. - Expires January 2006 23
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
whether the LSP is torn down by using a PathTear message (for an
egress-initiated LSP) or by using a PathErr message with the
Path_State_Removed flag set (for an ingress-initiated LSP).
If the Call was established using independent Call/Connection setup
and the Call is "empty" it may be deleted by the egress sending a
Notify message just as described above.
Note that there is still a possibility that both ends of a Call
initiate a simultaneous Call deletion. In this case, the Notify
message acting as teardown request is interpreted by its recipient as
a teardown response. Since the Notify messages carry the R bit in the
ADMIN STATUS object, they are responded to anyway. If a teardown
request Notify message is received for an unknown Call ID it is,
nevertheless, responded to in the affirmative.
7.8 Control Plane Survivability
Delivery of Notify messages is secured using message ID
acknowledgements as described in previous sections.
Notify messages provide end-to-end communication that does not rely
on constant paths through the network but are routed according to IGP
routing information. No consideration is, therefore, required for
network resilience (for example, make-before-break, protection, fast
re-route), although end-to-end resilience is of interest for node
restart and completely disjoint networks.
Periodic Notify messages SHOULD be sent by the initiator and
terminator of the Call to keep the Call alive and to handle ingress
or egress node restart. The time period for these retransmissions is
a local matter, but it is RECOMMENDED that this period should be
twice the refresh period of the LSPs associated with the Call. The
Notify messages are identical to those sent as if establishing the
Call for the first time, except for the LINK CAPABILITY object, which
may have changed since the Call was first established, due to, e.g.,
the establishment of connections, link failures, and the addition of
new component links. The current link information is useful for the
establishment of subsequent connections. A node that receives a
refresh Notify message MUST respond with a Notify response. A node
that receives a refresh Notify message (response or request) MAY
reset its timer - thus, in normal processing, Notify refreshes
involve a single exchange once per time period.
A node that is unsure of the status of a Call MAY immediately send a
Notify message as if establishing the Call for the first time.
Failure to receive a refresh Notify request has no specific meaning.
If it receives no response to a refresh Notify request (including no
Message ID Acknowledgement) a node MAY assume that the remote node is
unreachable or unavailable. It is a local policy matter whether this
causes the local node to teardown associated LSPs and delete the
Call.
D.Papadimitriou et al. - Expires January 2006 24
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
In the event that an edge node restarts without preserved state, it
MAY relearn LSP state from adjacent nodes and Call state from remote
nodes. If a Path or Resv message is received with a non-zero Call ID
but without the C bit in the ADMIN STATUS, and for a Call ID that is
not recognized, the receiver is RECOMMENDED to assume that the Call
establishment is delayed and ignore the received message. If the Call
setup never materializes the failure by the restarting node to
refresh state will cause the LSPs to be torn down. Optionally, the
receiver of such an LSP message for an unknown Call ID may return an
error (PathErr or ResvErr) with the error code "Call Management"
(TBD) and Error Value "Unknown Call ID" (TBD).
8. Applicability of Call and Connection Procedures
This section considers the applicability of the different Call
establishment procedures at the NNI and UNI reference points. This
section is informative and is not intended to prescribe or prevent
other options.
8.1 Network-initiated Calls
Both independent and simultaneous Call/Connection setups are
applicable.
Since the link properties and other traffic-engineering attributes
are likely known through the IGP, the LINK CAPABILITY object is not
usually required.
In multi-area networks, possibly, access link properties and other
traffic-engineering attributes are not known since the areas do not
leak this sort of information. In this case, the independent Call
setup mechanism may be preferred to allow the inclusion of the LINK
CAPABILITY object.
8.2 User-initiated Calls
Both independent and simultaneous Call/Connection setups are
applicable.
It is possible that the access link properties and other traffic-
engineering attributes are not shared across the core network. In
this case, the independent Call setup mechanism may be preferred to
allow the inclusion of the LINK CAPABILITY object.
Further, the first node in the network may be responsible for
managing the Call. In this case, the Notify message that is used to
set up the Call is addressed to the first node of the core network.
Moreover, neither the long Call ID nor the short Call ID is supplied
(the Session Name Length is set to zero and the Call ID value is set
to zero). The Notify message is passed to the first network node
which is responsible for generating the long and short Call IDs
before dispatching the message to the remote Call end point (which is
D.Papadimitriou et al. - Expires January 2006 25
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
known from the SESSION object). Similarly, the first network node may
be responsible for generating the long and short Call IDs for
inclusion in Path messages that have the C bit set in the ADMIN
STATUS object.
Further, when used in an overlay context, the first core node is
allowed (see [GMPLS-OVERLAY]) to replace the Session Name assigned by
the ingress node and passed in the Path message. In the case of Call
management, the first network node MUST in addition 1) be aware that
the name it inserts MUST be a long Call ID and 2) replace the long
Call ID when it returns the Resv message to the ingress node.
8.3 External Call Managers
Third party Call management agents may be used to apply policy and
authorization at a point that is neither the initiator nor terminator
of the Call. The previous example is a particular case of this, but
the process and procedures are identical.
8.3.1 Call Segments
Call segments exist between a set of default and configured External
Call Managers along a path between the ingress and egress nodes, and
use the protocols described in this document.
The techniques that are used by a given service provider to identify
which External Call Managers within its network should process a
given call are beyond the scope of this document.
For independent call setup, an External Call manager uses normal IP
routing to route the Notify message to the next External Call
Manager. For simultaneous call/connection setup, an External Call
Manager expands the EXPLICIT_ROUTE Object to route the Path message
to the next External Call Manager.
9. Non-support of Call ID
It is important that the procedures described above operate as
seamlessly as possible with legacy nodes that do not support the
extensions described.
Clearly there is no need to consider the case where the Call
initiator does not support Call setup initiation.
9.1 Non-Support by External Call Managers
It is unlikely that a Call initiator will be configured to send Call
establishment Notify requests to an external Call manager including
the first network node, if that node does not support Call setup.
A node that receives an unexpected Call setup request will fall into
one of the following categories.
D.Papadimitriou et al. - Expires January 2006 26
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
- Node does not support RSVP. The message will fail to be delivered
or responded. No Message ID Acknowledgement will be sent. The
initiator will retry and then give up.
- Node supports RSVP or RSVP-TE but not GMPLS. The message will be
delivered but not understood. It will be discarded. No Message ID
Acknowledgement will be sent. The initiator will retry and then
give up.
- Node supports GMPLS but not Call management. The message will be
delivered, but parsing will fail because of the presence of the
SESSION ATTRIBUTE object. A Message ID Acknowledgement may be sent
before the parse fails. When the parse fails the Notify message
may be discarded in which case the initiator will retry and then
give up, alternatively a parse error may be generated and returned
in a Notify message which will indicate to the initiator that Call
management is not set up.
9.2 Non-Support by Transit Node
Transit nodes SHOULD not examine Notify messages that are not
addressed to them. However, they will see short Call IDs in all LSPs
associated with Calls. Further, they will see the C bit in the ADMIN
STATUS object of Path and Resv messages that are used to establish
Calls.
Previous specifications state that these fields SHOULD be ignored on
receipt and MUST be transmitted as zero. This is interpreted by some
implementations as meaning that the fields should be zeroed before
the objects are forwarded. If this happens, LSP setup (and so
possibly Call setup if simultaneous establishment is used) will not
be possible. If either of the fields is zeroed either on the Path or
the Resv message, the Resv will reach the initiator with the field
set to zero - this is indication to the initiator that some node in
the network is preventing Call management. Use of Explicit Routes may
help to mitigate this issue by avoiding such nodes. The use of
independent Call setup may also help since it removes the need for
the C bit in the Path and Resv messages. Ultimately, however, it may
be necessary to upgrade the offending nodes to handle these protocol
extensions.
9.3 Non-Support by Egress Node
It is unlikely that an attempt will be made to set up a Call to
remote node that does not support Calls.
If the egress node does not support Call management through the
Notify message it will react (as described in Section 9.1) in the
same way as an external Call manager.
If the egress node does not support the use of the C bit in the ADMIN
STATUS object or the Call ID in the SESSION object, it MAY respond
D.Papadimitriou et al. - Expires January 2006 27
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
with the fields zeroed in which case the initiator will know that the
Call setup has failed.
On the other hand, it is possible that the egress will respond
copying the fields from the Path message without understanding or
acting on the fields. In this case the initiator will believe that
the Call has been set up when it has not. This occurrence can be
prevented using the independent Call setup procedures, but is, in any
case, detected when a Notify message is sent to keep the Call alive.
10. Security Considerations
Please refer to each of the referenced documents for a description of
the security considerations applicable to the features that they
provide.
10.1 Call and Connection Security Considerations
Call setup is vulnerable to attacks both of spoofing and denial of
service. Since Call setup uses either Path messages or Notify
messages, the process can be protected by the measures applicable to
securing those messages as described in [RFC3471], [RFC3209] and
[RFC2205].
Note, additionally, that the process of Call establishment
independent of LSP setup may be used to apply an extra level of
authentication and policy to hop-by-hop LSP setup. It may be possible
to protect the Call setup exchange using end-to-end security
mechanisms such as those provided by Insect (see [RFC2402] and
[RFC2406]).
11. IANA Considerations
A new RSVP object is introduced:
o LINK CAPABILITY object
Class-Num = TBA (form 10bbbbbb)
The Class Number is selected so that nodes not recognizing
this object drop it silently. That is, the top bit is set
and the next bit is cleared.
C-Type = 1 (TE Link Capabilities)
New RSVP Error Codes and Error Values are introduced
o Error Codes:
- Call Management (value TBA)
o Error Values:
D.Papadimitriou et al. - Expires January 2006 28
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
- Call Management/Call ID Contention (value TBA)
- Call Management/Connections still Exist (value TBA)
- Call Management/Unknown Call ID (value TBA)
12. Acknowledgements
The authors would like to thank George Swallow, Yakov Rekhter, Lou
Berger, Jerry Ash and Kireeti Kompella for their very useful input
and comments to this document.
13. References
13.1 Normative References
[ASON-REQ] D.Papadimitriou, et al., "Requirements for
Generalized MPLS (GMPLS) Usage and Extensions for
Automatically Switched Optical Network (ASON)," Work
in progress, Oct'04.
[BUNDLE] K.Kompella, Y.Rekhter and L.Berger, "Link Bundling
in MPLS Traffic Engineering," Work in Progress.
[GMPLS-CRANK] A.Farrel (Editor) et al., "Crankback Routing
Extensions for MPLS Signaling," Work in progress,
Oct'04.
[GMPLS-FUNCT] J.P.Lang and B.Rajagopalan (Editors) et al.,
"Generalized MPLS Recovery Functional
Specification," Work in Progress, Oct'04.
[GMPLS-OVERLAY] G.Swallow et al., "GMPLS RSVP Support for the
Overlay Model," Work in Progress, Oct'04.
[GMPLS-ROUTING] K.Kompella and Y.Rekhter (Editors) et al., "Routing
Extensions in Support of Generalized MPLS," Work in
Progress, Oct'03.
[LMP] J.P.Lang (Editor) et al. "Link Management Protocol
(LMP) - Version 1," Work in progress, Oct'03.
[RFC2026] S.Bradner, "The Internet Standards Process --
Revision 3," BCP 9, RFC 2026, Oct'96.
[RFC2119] S.Bradner, "Key words for use in RFCs to Indicate
Requirement Levels," BCP 14, RFC 2119, Mar'97.
[RFC2205] R.Braden et al., "Resource ReSerVation Protocol
(RSVP)- Version 1 Functional Specification,"
RFC 2205, Sep'97
[RFC2402] S.Kent and R.Atkinson, "IP Authentication Header,"
RFC 2402, Nov'98.
D.Papadimitriou et al. - Expires January 2006 29
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
[RFC2406] S.Kent and R.Atkinson, "IP Encapsulating Payload
(ESP)," RFC 2406, Nov'98.
[RFC3209] D.Awduche et al., "RSVP-TE: Extensions to RSVP for
LSP Tunnels," RFC 3209, Dec'01.
[RFC3471] L.Berger (Editor) et al., "Generalized MPLS -
Signaling Functional Description," RFC 3471, Jan'03.
[RFC3473] L.Berger (Editor) et al., "Generalized MPLS
Signaling - RSVP-TE Extensions," RFC 3473, Jan'03.
[RFC3477] K.Kompella and Y.Rekhter, "Signalling Unnumbered
Links in Resource ReSerVation Protocol - Traffic
Engineering (RSVP-TE)," RFC 3477, Jan'03.
[RFC3667] S.Bradner, "IETF Rights in Contributions", BCP 78,
RFC 3667, February 2004.
[RFC3668] S.Bradner, Ed., "Intellectual Property Rights in IETF
Technology", BCP 79, RFC 3668, February 2004.
[RFC3945] E.Mannie, Ed., "Generalized Multi-Protocol Label
Switching (GMPLS) Architecture", RFC 3945, October
2004.
[RSVP-CHANGE] K.Kompella and J.P.Lang, "Procedures for Modifying
RSVP," Work in Progress, draft-kompella-rsvp-change-
01.txt, Jun'03.
13.2 Informative References
[RFC3474] Z.Lin (Editor), " Documentation of IANA assignments
for Generalized MultiProtocol Label Switching
(GMPLS) Resource Reservation Protocol - Traffic
Engineering (RSVP-TE) Usage and Extensions for
Automatically Switched Optical Network (ASON)," RFC
3474, Mar'03.
[RFC3476] B.Rajagopalan (Editor), "Documentation of IANA
Assignments for Label Distribution Protocol
(LDP), Resource ReSerVation Protocol (RSVP), and
Resource ReSerVation Protocol-Traffic Engineering
(RSVP-TE) Extensions for Optical UNI Signaling," RFC
3476, Mar'03.
For information on the availability of the following documents,
please see http://www.itu.int.
[G.7713] ITU-T, "Distributed Call and Connection Management,"
Recommendation G.7713/Y.1304, Nov'01.
D.Papadimitriou et al. - Expires January 2006 30
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
[G.7713.2] ITU-T, "DCM Signalling Mechanisms Using GMPLS RSVP-
TE," Recommendation G.7713.2, Jan'03.
[G.8080] ITU-T, "Architecture for the Automatically Switched
Optical Network (ASON)," Recommendation G.8080/
Y.1304, Nov'01 (and Revision, Jan'03).
14. Author's Addresses
Dimitri Papadimitriou (Alcatel)
Fr. Wellesplein 1,
B-2018 Antwerpen, Belgium
Phone: +32 3 240-8491
EMail: dimitri.papadimitriou@alcatel.be
John Drake
Boeing Satellite Systems
2300 East Imperial Highway
El Segundo, CA 90245
EMail: John.E.Drake2@boeing.com
Adrian Farrel
Old Dog Consulting
Phone: +44 (0) 1978 860944
EMail: adrian@olddog.co.uk
Deborah Brungard (AT&T)
Rm. D1-3C22 - 200 S. Laurel Ave.
Middletown, NJ 07748, USA
EMail: dbrungard@att.com
Zafar Ali (Cisco)
100 South Main St. #200
Ann Arbor, MI 48104, USA
EMail: zali@cisco.com
Arthi Ayyangar (Juniper)
1194 N.Mathilda Ave
Sunnyvale, CA 94089, USA
EMail: arthi@juniper.net
Don Fedyk (Nortel Networks)
600 Technology Park Drive
Billerica, MA, 01821, USA
Email: dwfedyk@nortelnetworks.com
D.Papadimitriou et al. - Expires January 2006 31
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
Appendix 1: Analysis of G.7713.2 against GMPLS RSVP-TE Signaling
Requirements in support of ASON
Informational RFC [RFC3474] (and [RFC3476]) documents the code points
for the signaling extensions defined in [G.7713.2] to meet the
requirements of ASON Distributed Call and Connection Management (DCM)
as specified in [G.7713].
While [G.7713.2] make use of GMPLS RSVP-TE signaling, there are key
differences from the problem statement in [ASON-REQ] and the solution
it provides. These differences result from the development of a
fuller and clearer set of requirements in [G.8080] after the time
that [G.7713.2] was published and [ASON-REQ] considerations for
compatibility aspects with GMPLS [RFC3473]. These differences lead to
a substantially different protocol solution and implementation.
This appendix analyzes the rationale and the relevance of the
informational IANA code-point assignments RFCs [RFC3474] and
[RFC3476] against the ASON requirements identified in [ASON-REQ]. The
latter details the requirements to be covered by the extensions to
the GMPLS signaling protocols (see [RFC3471] and [RFC3473]) to
support the capabilities of an ASON network. The following are
expected from the GMPLS protocol suite to realize the needed ASON
functionality:
o soft permanent connection capability
o call and connection separation
o call segments
o extended restart capabilities during control plane failures
o extended label usage
o crankback capability
1. Support for UNI and E-NNI Signaling Session
In GMPLS (see [RFC3473] and related), a connection is identified with
a GMPLS tunnel. A tunnel is generally identified with a single LSP
but may be supported by multiple LSPs.
LSP tunnels are identified by a combination of the SESSION and
SENDER_TEMPLATE objects. The relevant fields are as follows.
IPv4 (or IPv6) tunnel end point address
IPv4 (or IPv6) address of the egress node for the tunnel.
Tunnel ID
A 16-bit identifier used in the SESSION that remains constant
over the life of the tunnel.
Extended Tunnel ID
D.Papadimitriou et al. - Expires January 2006 32
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
A 32-bit (IPv4) or 128-bit (IPv6) identifier used in the SESSION
that remains constant over the life of the tunnel. Normally set
to all zeros. Ingress nodes that wish to narrow the scope of a
SESSION to the ingress-egress pair may place their IP address
here as a globally unique identifier.
IPv4 (or IPv6) tunnel sender address
IPv4 (or IPv6) address for a sender node
LSP ID
A 16-bit identifier used in the SENDER_TEMPLATE and the
FILTER_SPEC that can be changed to allow a sender to share
resources with itself.
The first three of these are in the SESSION object and are the basic
identification of the tunnel. The "Extended Tunnel ID" MAY be set to
an IP address of the head-end LSR allowing the scope of the SESSION
to be narrowed to only LSPs sent by that node. The last two are in
the SENDER_TEMPLATE. Multiple LSPs may belong to the same tunnel (and
thus SESSION) and in this case they are uniquely identified by their
LSP IDs.
In contrast, [G.7713.2] defines an E-NNI IPv4/IPv6 SESSION object and
an UNI IPv4/IPv6 SESSION object. It mandates the use of these objects
to support the E-NNI (UNI, respectively) signaling session when IPv4
and IPv6 addressing is used. The "Tunnel End-point Address" field
contains the IPv4 or IPv6 address of the downstream controller. In
addition, [G.7713.2] mandates that the "Extended Tunnel ID" field to
be set to the IPv4 or IPv6 of the upstream controller. It also
mandates that the tunnel sender address field of the SENDER_TEMPLATE
be set to the IPv4 or the IPv6 address of the upstream controller.
Thus, these RFCs define a point-to-point signaling interface allowing
for LSP tunnel provisioning between adjacent controllers only. This
approach mandates the introduction of an additional object and sub-
objects for connection identification purposes (see [G.7713.2]): the
GENERALIZED_UNI object and its connection end-point address sub-
objects (IPv4/IPv6/NSAP) referred to as "TNA or Transport Network
Address" as defined by the [OIF-UNI] implementation agreement.
The situation is summarized in the following table, using the
following example and a connection set from node A to Z:
UNI E-NNI E-NNI UNI
A ----- B -- ... -- I ----- J -- .. -- M ----- N -- ... -- Y ----- Z
At node I, the GMPLS compliant [RFC3473] Path message would include
the following information in the IP header, the SESSION and
SENDER_TEMPLATE objects:
D.Papadimitriou et al. - Expires January 2006 33
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
Source IP address (Header): Node I IP Controller Address
Dest. IP address (Header): Node J IP Controller Address
Tunnel End-point Address: Routable Node Z IP Address
Tunnel ID: 16 bit (selected by the sender)
Extended Tunnel ID: optionally set to Node A IP Address
Tunnel Sender Address: Routable Node A IP Address
LSP ID: 16 bit (selected by the sender)
At node I, the [G.7713.2] Path message would include the following:
Source IP address (Header): Node I IP Controller Address
Dest. IP address (Header): Node J IP Controller Address
Tunnel End-point Address: Node J IP Controller Address
Tunnel ID: 16 bit (selected by the sender)
Extended Tunnel ID: Node I IP Controller Address
Tunnel Sender Address: Node I IP Controller Address
LSP ID: 16 bit (selected by the sender)
GENERALIZED_UNI object:
- Source Address (Connection): End-point A Address (IPv4/IPv6/NSAP)
- Dest. Address (Connection): End-point Z Address (IPv4/IPv6/NSAP)
The same observation would apply at node M, by replacing I by M and J
by N.
The following can be thus deduced from the above example:
1. For a given connection, the [G.7713.2] point-to-point signaling
interface leads to a sequence of at least N different
identifications of the same connection when crossing N
signaling interfaces (due to the setup and maintenance of N
distinct LSP tunnels).
2. The information included in the RSVP message header and in the
SESSION/SENDER_TEMPLATE objects, is redundant in [G.7713.2].
3. [G.7713.2] allows only for single-hop LSP tunnels and mandates
the processing of a new object, i.e. the GENERALIZED_UNI object,
for the definition of the source and destination connection end-
point addresses (A and Z in the above example).
4. The processing of the signaling Path message, in particular, the
EXPLICIT ROUTE object (ERO), mandates the processing of the
GENERALIZED_UNI object at E-NNI reference points and at UNI
reference points, for the connection end-point addresses (A and
Z, in the above example).
5. Connection end-point addresses A and Z are allowed by [G.7713.2]
to be from different address spaces (for instance, IPv4 source
and IPv6 destination or an IPv4 source and NSAP destination).
Address resolution, management of addresses, e.g., for
uniqueness, and impact evaluation on processing performance, are
not provided in these RFCs (considered out of scope).
D.Papadimitriou et al. - Expires January 2006 34
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
Note: the [ASON-REQ] addressing model of supporting only IP
addressing is aligned with ITU-T G.7713.1 (PNNI) which only uses
NSAP addresses, multiple address families are not supported.
6. [G.7713.2] defines an incompatible and redundant addressing
mechanism with [RFC3473] to support IPv4, IPv6, and NSAP
addresses. [RFC3473] is part of a GMPLS protocol suite based on
use of IPv4 and IPv6 addresses. The use of NSAP addresses with
[RFC3473] is supported by established procedures defined in
[RFC1884] "IPv6 Addressing Architecture", and only requiring
support by border entities, e.g., DNS. Any other support for
NSAP addressing is redundant with IETF supported procedures.
[G.7713.2] provides no guidance on the operational aspects
resulting from this modified procedure which uses a non-standard
object, the GENERALIZED_UNI object, to support. Use of the
GENERALIZED_UNI object requires every entity to support multi-
address family resolution, e.g., for ERO processing, and in the
case of multi-region path setup. Requiring multi-address family
resolution at all entities severely impacts performance, scaling,
and introduces unnecessary complexity for operations. This
limitation is well recognized, e.g. [G.7713.2] use in demos has
been limited to only IPv4 prefixes with pre-configured mappings.
Conclusion:
1) The solution proposed by [G.7713.2] is not backward compatible
with [RFC3473]. A GMPLS-compliant node [RFC3473] is not interoperable
with a [G.7713.2] node. Also, the "RSVP paradigm" is broken because
the solution requires that all the UNI reference points (A, B and Y,
Z, in the above example) and the E-NNI reference points (I, J and M,
N, in the above example) support the GENERALIZED_UNI object.
Additionally, the management of the network requires maintaining
multiple LSP tunnels per single connection, with no end-to-end view
provided for expedient fault notification or recovery operations.
2) The solution proposed by [G.7713.2] also introduces processing
overhead for address resolution that during time critical operations
(such as recovery) will severely impact performance and scalability.
Whereas the ITU-T G.7713.1 (PNNI) and [ASON-REQ] by using a single
address family (with address mapping provided at edge nodes if
needed) supports a scalable model for inter-domain interworking
applications.
2. Support for Soft Permanent Connections (SPC)
A Soft Permanent Connection (SPC) is defined as a permanent
connection at the network edges and as a switched connection within
the network.
[G.7713.2] mandates the use of the GENERALIZED_UNI subobjects (End-
point Connection Address and SPC_LABEL) to support SPC capability.
Thus, in addition to suffering from the problem described in Section
4, it mandates the processing of an object where it should never
D.Papadimitriou et al. - Expires January 2006 35
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
occur. This is because SPCs do not assume the existence of a UNI
signaling interface between the source and the destination user-to-
network interface. Note also that the SPC_LABEL as defined in
[G.7713.2] does not comply with the generalized label C-Type
definition of [RFC3473] meaning that an implementation adhering to
[RFC3473] cannot be the "soft" side of the connection.
This requires the mandatory usage of dedicated connection end-point
addresses by the ingress and egress nodes for SPC capability support.
The existing RECORD_ROUTE object and its capabilities get corrupted
by the use of the dedicated end-point address subobjects falling
outside of the corresponding EXPLICIT ROUTE object.
SPC support is already provided by [RFC3473] using Explicit Label
Control and its application to the overlay model in [GMPLS-OVERLAY].
Therefore, [G.7713.2] defines a new method for an existing capability
of GMPLS signaling.
3. Call/Connection Separation
The call concept for optical networks is defined in [G.8080]. It is
used to deliver the following capabilities:
- Verification and identification of the call initiator (prior to
LSP setup) including negotiation between call ingress/egress nodes
- Support of multiple connections can be associated with a single
call.
- Facilitate control plane operations by allowing operational status
change of the associated LSP.
A call is an agreement between end-points (possibly in cooperation
with the nodes that provide access to the network) used to manage a
set of connections that provide end to end services. While
connections require state to be maintained at nodes along the data
path within the network, *** calls do not involve the participation
of transit nodes except to forward the call management requests as
transparent messages ***. Moreover, a call may be established and
maintained independently of the connections that it supports.
Also, there is a hierarchical relationship between calls and
connections. One or more (or even no) connections may be associated
with a given call but a connection can not be part of more than one
call. A connection may, however, exist without a call. Moreover, the
establishment of a call can be independent ("full call/connection
separation") or simultaneous ("logical call/connection separation")
from the connection setup (i.e. establishing a call before adding
connections to the call or perform these operations simultaneously).
Thus, with the introduction of the call concept, it is necessary to
support a means of identifying the call. This becomes important when
calls and connections are separated and a connection must contain a
reference to its associated call. The following identification
enables this hierarchy:
- Call IDs are unique within the context of the pair of addresses
D.Papadimitriou et al. - Expires January 2006 36
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
that are the source and destination of the call.
- Tunnel IDs are unique within the context of the Session (that is
the destination of the Tunnel) and Tunnel IDs may be unique within
the context of a Call.
- LSP IDs are unique within the context of a Tunnel.
For this purpose, [G.7713.2] introduces two new objects: a CALL_ID
and a CALL_OPS object to be used in the Path, Resv, PathTear,
PathErr, and Notify messages (note: additional requirements for
ResvErr and ResvTear messages' support are not addressed). The
CALL_OPS object is also referred to as a "call capability" object,
since it specifies the capability of the call. These objects belongs
to the range 224-255 defined as "RSVP will silently ignore, but
FORWARD an object with a Class Number in this range that it does not
understand."
However, the solution described in [G.7713.2]:
- Does not provide backward compatible extensions in support of full
call/connection separation and thus only supports logical call/
connection separation (i.e. a call with zero connections is not
supported). This because node that does not implement [G.7713.2],
will not process the CALL_OPS object, though it will establish the
*connection* (while forwarding the "Call Setup" message), i.e.
allocating labels and possibly attempting to reserve bandwidth.
[G.7713.2] forbids this behavior by a transit node, but only a
node implementing [G.7713.2] will know the difference between a
call and a connection.
In turn, the required signaling protocol independence between
intra- and inter-domain reference points is broken: an operator
does not have the possibility to use GMPLS [RFC3473] and must
exclusively use [G.7713.2].
- Does not describe how to support multiple connections per call but
limits the description to a single connection per call. Further,
in the case of complete call/connection separation, it does not
describe how to add the first connection to the call.
- Does not describe how to support multiple connections per call and
limits the description to a single connection per call. Further,
it does not describe how to add the first connection to the call
when to support call/connection separation.
- Does not specify any procedure for negotiating call ingress/egress
node capabilities during call setup.
- Does not allow for call support *independently* of the initiating/
terminating nodes (the CALL_ID is attached to the ingress node)
thus restricting the flexibility in terms of call identifiers.
- Requires the inclusion of the CALL ID and CALL OPS objects in
PathErr messages that may be generated at transit nodes, which do
D.Papadimitriou et al. - Expires January 2006 37
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
not implement [G.7713.2] and so do not support these objects.
4. Call Segments
[G.7713.2] cannot, by definition, support call segments signaling
mechanisms, as required in [G.8080] and [G.7713], since [G.7713.2]
does not support full call/connection separation.
5. Control Plane Restart Capabilities
Restart capabilities are provided by GMPLS RSVP-TE signaling in case
of control plane failure including nodal and control channel faults.
The handling of node and control channels faults is described in
[RFC3473] Section 9. No additional RSVP mechanisms or objects are
required to fulfill the ASON control plane restart capabilities.
However, [G.7713.2] defines additional procedures for control plane
recovery, three of them being considered in the context of an
interaction with the management plane and thus outside the scope of
the present document. The last one expects persistent state storage
and the restart mechanism defined in [RFC3473] is to be used for
verification of neighbor states, while the persistent storage
provides the local recovery of lost state. However, per [RFC3473], if
during the Hello synchronization the restarting node determines that
a neighbor does not support state recovery and the restarting node
maintains its local state on a per neighbor basis, the restarting
node should immediately consider the Recovery as completed. Therefore,
the procedure described in [G.7713.2] requires disabling state
recovery on each neighboring node leading also to an unspecified
verification procedure.
6. Extended Label Usage
No specific GMPLS RSVP-TE extensions have been proposed in [G.7713.2]
for extended label usage.
7. Crankback Signaling
[G.7713.2] does not support crankback signaling mechanisms, as
required in [G.8080] and [G.7713].
8. Security Considerations
This is an informational draft and does not introduce any new
security considerations.
Please refer to each of the referenced documents for a description of
the security considerations applicable to the features that they
provide.
Note that although [RFC3474] is an informational RFC it does document
new protocol elements and functional behavior and as such introduces
new security considerations. In particular, the ability to place
D.Papadimitriou et al. - Expires January 2006 38
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
authentication and policy details within the context of Call
establishment may strengthen the options for security and may weaken
the security of subsequent Connection establishment. Also the
potential to subvert accidentally or deliberately a soft permanent
connection by establishing the soft part of the connection from a
false remote node needs to be examined. However, [RFC3474] has a
minimal security considerations section.
D.Papadimitriou et al. - Expires January 2006 39
draft-ietf-ccamp-gmpls-rsvp-te-ason-04.txt July 2005
Intellectual Property Statement
The IETF takes no position regarding the validity or scope of any
Intellectual Property Rights or other rights that might be claimed to
pertain to the implementation or use of the technology described in
this document or the extent to which any license under such rights
might or might not be available; nor does it represent that it has
made any independent effort to identify any such rights. Information
on the procedures with respect to rights in RFC documents can be
found in BCP 78 and BCP 79.
Copies of IPR disclosures made to the IETF Secretariat and any
assurances of licenses to be made available, or the result of an
attempt made to obtain a general license or permission for the use of
such proprietary rights by implementers or users of this
specification can be obtained from the IETF on-line IPR repository at
http://www.ietf.org/ipr.
The IETF invites any interested party to bring to its attention any
copyrights, patents or patent applications, or other proprietary
rights that may cover technology that may be required to implement
this standard. Please address the information to the IETF at ietf-
ipr@ietf.org.
Disclaimer of Validity
This document and the information contained herein are provided on an
"AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE
INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Copyright Statement
Copyright (C) The Internet Society (2005). This document is subject
to the rights, licenses and restrictions contained in BCP 78, and
except as set forth therein, the authors retain all their rights.
Acknowledgment
Funding for the RFC Editor function is currently provided by the
Internet Society.
D.Papadimitriou et al. - Expires January 2006 40