Proxy Mobile IPv6 (PMIPv6) Localized Routing Problem Statement
Routing data directly between access gateways instead of through a distant anchor.
Liebsch, M., Ed., Jeong, S., and Q. Wu, "Proxy Mobile IPv6 (PMIPv6) Localized Routing Problem Statement", RFC 6279, DOI 10.17487/RFC6279, June 2011, <https://www.rfc-editor.org/info/rfc6279>.
@misc{rfc6279,
series = {Request for Comments},
number = 6279,
howpublished = {RFC 6279},
publisher = {RFC Editor},
doi = {10.17487/RFC6279},
url = {https://www.rfc-editor.org/info/rfc6279},
author = {M. Liebsch and S. Jeong and Q. Wu},
title = {{Proxy Mobile IPv6 (PMIPv6) Localized Routing Problem Statement}},
pagetotal = 14,
year = 2011,
month = jun,
} RFC text
Contents
Internet Engineering Task Force (IETF) M. Liebsch, Ed.
Request for Comments: 6279 NEC
Category: Informational S. Jeong
ISSN: 2070-1721 ETRI
Q. Wu
Huawei
June 2011
Proxy Mobile IPv6 (PMIPv6) Localized Routing Problem Statement
Abstract
Proxy Mobile IPv6 is the IETF Standard for network-based mobility
management. In Proxy Mobile IPv6, mobile nodes are topologically
anchored at a Local Mobility Anchor, which forwards all data for
registered mobile nodes. The setup and maintenance of localized
routing, which allows forwarding of data packets between two mobile
nodes' Mobility Access Gateways without involvement of their Local
Mobility Anchor in forwarding, is not considered. This document
describes the problem space of localized routing in Proxy Mobile
IPv6.
Status of This Memo
This document is not an Internet Standards Track specification; it is
published for informational purposes.
This document is a product of the Internet Engineering Task Force
(IETF). It represents the consensus of the IETF community. It has
received public review and has been approved for publication by the
Internet Engineering Steering Group (IESG). Not all documents
approved by the IESG are a candidate for any level of Internet
Standard; see Section 2 of RFC 5741.
Information about the current status of this document, any errata,
and how to provide feedback on it may be obtained at
http://www.rfc-editor.org/info/rfc6279.
Copyright Notice
Copyright (c) 2011 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Table of Contents
1. Introduction ....................................................2
2. Conventions and Terminology .....................................3
3. Problem Statement for Localized Routing in PMIPv6 ...............4
3.1. General Observation ........................................4
3.2. Use Cases Analysis .........................................5
3.3. IPv4 Considerations ........................................8
3.3.1. Localized Routing for Communication between
IPv4 Home Addresses .................................8
3.3.2. IPv4 Transport Network Considerations ...............9
4. Functional Requirements for Localized Routing ...................9
5. Roaming Considerations .........................................10
6. Security Considerations ........................................11
7. Acknowledgments ................................................12
8. References .....................................................13
8.1. Normative References ......................................13
8.2. Informative References ....................................13
1. Introduction
The IETF has specified Proxy Mobile IPv6 (PMIPv6) [RFC5213] as the
base protocol for network-based localized mobility management
(NetLMM). The scope of the base protocol covers the setup and
maintenance of a tunnel between a Mobile Node's (MN's) Mobile Access
Gateway (MAG) and its selected Local Mobility Anchor (LMA). Data
packets will always traverse the MN's MAG and its LMA, irrespective
of the location of the MN's remote communication endpoint. Even
though an MN may be attached to the same MAG or a different MAG as
its Correspondent Node (CN) within the same provider domain, packets
being associated with their communication will traverse the MN's and
the CN's LMA, which can be located topologically far away from the
MN's and the CN's MAG or even in a separate provider domain.
[RFC5213] addresses the need to enable local routing of traffic
between two nodes being attached to the same MAG, but does not
specify the complete procedure to establish such localized routing
state on the shared MAG.
The NetLMM Extensions (NetExt) Working Group has an objective to
design a solution for localized routing in PMIPv6. This objective
includes the specification of protocol messages and associated
protocol operation between PMIPv6 components to support the setup of
a direct routing path for data packets between the MN's and the CN's
MAG, while both hosts receive mobility service according to the
PMIPv6 protocol [RFC5213]. As a result of localized routing, these
packets will be forwarded between the two associated MAGs without
traversing the MN's and the CN's LMA(s). In cases where one or both
nodes hand over to a different MAG, the localized routing protocol
maintains the localized routing path. Relevant protocol interfaces
may include the interface between associated MAGs, between a MAG and
an LMA, and between LMAs. The setup of localized routing with CNs
not registered with a PMIPv6 network is out of scope of the NetExt
solution and this problem statement.
This document analyzes and discusses the problem space of always
using the default route through two communicating mobile nodes' local
mobility anchors. Furthermore, the problem space of enabling
localized routing in PMIPv6 is analyzed and described, while
different communication and mobility scenarios are taken into
account. Based on the analysis, a list of key functional
requirements is provided, serving as input to the design of the
protocol solution.
2. Conventions and Terminology
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].
This document uses the terminology of [RFC5213]. In addition, the
following terms are used in the context of this problem statement:
o Mobile Node (MN): Mobile Node without IP mobility support, which
is attached to a Mobile Access Gateway (MAG) and registered with a
Local Mobility Anchor (LMA) according to the PMIPv6 specification
[RFC5213].
o Correspondent Node (CN): Correspondent Node according to its
definition in [RFC3775] with or without IP mobility support. The
CN represents the communication peer of an MN that is attached to
a MAG and registered with an LMA according to the PMIPv6
specification.
o Localized Routing: Result of signaling to set up routing states on
relevant network entities to allow forwarding of data packets
between an MN and a CN, which are attached to MAGs sharing the
same provider domain, without intervention of the MN's LMA and the
CN's LMA in data forwarding.
o Localized Routing States: Information for localized routing on
relevant forwarding entities on the optimized data path between an
MN and a CN. Such information includes route entries and tunnel
endpoints and may include further information about the MN and the
CN, such as the communicating nodes' Mobile Node Identifier and
their assigned Home Network Prefix.
o Provider Domain: A network domain in which network components are
administered by a single authority, e.g., the mobile operator.
3. Problem Statement for Localized Routing in PMIPv6
3.1. General Observation
The Mobile IPv6 (MIPv6) protocol [RFC3775] has built-in mechanisms
for direct communication between an MN and a CN. Mechanisms for
route optimization in MIPv6 cannot be directly applied in PMIPv6.
Following the paradigm of PMIPv6, MNs are not involved in mobility
signaling and hence cannot perform signaling to set up localized
routes. Instead, the solution for localized routing must consider
functions in the network to find out whether or not a localized route
is to be used and then control the setup and maintenance of localized
routing states accordingly without any assistance from the MN and the
CN. In the case of communication between two nodes attached to the
PMIPv6 network infrastructure and where each node is registered with
an LMA, data packets between these two nodes will always traverse the
responsible LMA(s). At least some deployments would benefit from
having such communication localized, rather than having packets
traverse the core network to the LMA(s). In the context of this
document, such localized communication comprises offloading traffic
from LMAs and establishing an optimized forwarding path between the
two communication endpoints.
Localized routing is understood in [RFC5213] as optimization of
traffic between an MN and a CN that are attached to an access link
connected to the same MAG. In such a case, the MAG forwards traffic
directly between the MN and the CN, assuming the MAG is enabled to
support this feature (setting of the EnableMAGLocalRouting flag on
the MAG) and the MN's LMA enforces this optimization. [RFC5213] does
not specify how an LMA can enforce optimization for such local
communication. Maintaining local forwarding between the MN and the
regular IPv6 CN gets more complex in the case where the MN performs a
handover to a different MAG. Such a use case is not considered in
the specification and is out of scope of this problem statement.
This document focuses on use cases where both nodes, the MN and the
CN, are within a PMIPv6 network and served by an LMA in a domain of
LMAs.
With localized routing, operators have the possibility of offloading
traffic from LMAs and from the core network. Establishment of a
direct path between the MN's and the CN's MAG can be beneficial for
the following reasons: First, by limiting the communication to the
access nodes, the data traffic traversing the MAG - LMA path
(network) can be reduced. This is significant, considering that the
transport network between the access and the core is often the
bottleneck in terms of costs and performance. Second, there may be
performance benefits for data flows between the MN and the CN in
terms of delay and packet loss, especially when the MN and the CN are
attached to the same MAG and the LMA is topologically far away from
that MAG. Even when the MN and the CN are attached to different
MAGs, there could be benefit in limiting the communication to the
access network only, rather than traversing the transport network to
the LMA. Furthermore, offloading traffic from the LMA by means of
localized routing can improve scalability of the LMA, as it
represents a bottleneck for traffic being forwarded by many MAGs.
3.2. Use Cases Analysis
This problem statement focuses on local communication between PMIPv6
managed nodes, which attach to MAGs sharing the same provider domain.
The following list analyzes different use cases, which consider the
existence of multiple LMAs. Figure 1 depicts a PMIPv6-based network
with two mobility anchors. According to [RFC5213], the MN moves in
the PMIPv6 domain being built by its LMA and MAG. The same applies
to the CN, which moves in the PMIPv6 domain built by the CN's LMA and
MAG. The analysis takes no assumption on whether the MN and the CN
share the same PMIPv6 domain or not.
Internet Backbone
: :
+------------------+
| |
+----+ +----+
|LMA1| |LMA2|
+----+ +----+
| |
| |
+----+------------------+----+
| |
+----+ +----+
|MAG1| |MAG2|
+----+ +----+
: : :
+---+ +---+ +---+
|MN | |CN1| |CN2|
+---+ +---+ +---+
Figure 1: Reference Architecture for Localized Routing in PMIPv6
All "A" use cases below assume that both the MN and the CN are
registered with an LMA according to the PMIPv6 protocol. Whereas
MAG1 is always considered as the MN's current Proxy Care-of Address,
the CN can be either connected to the same MAG or to a different MAG
or LMA as the MN. Accordingly, these topological differences are
denoted as follows:
A[number of MAGs][number of LMAs]
A11: The MN and the CN (CN1) connect to the same MAG (MAG1) and are
registered with the same LMA (LMA). The common MAG may forward
data packets between the MN and the CN directly without forwarding
any packet to the LMA. [RFC5213] addresses this use case, but
does not specify the complete procedure to establish such
localized routing state on the shared MAG.
A12: The MN and the CN (CN1) connect to the same MAG (MAG1) and are
registered with different LMAs (LMA1 and LMA2). The common MAG
may forward data packets between the MN and the CN directly
without forwarding any packet to the LMAs. Following the policy
of [RFC5213] and enforcement of the setup of a localized
forwarding path, potential problems exist in the case where LMA1
and LMA2 differ in their policy to control the MAG.
A21: The CN (CN2) connects to a different MAG (MAG2) than the MN
(MAG1), but the MN and the CN are registered with the same LMA
(LMA1). The result of localized routing should be the existence
of routing information at MAG1 and MAG2, which allows direct
forwarding of packets between the MN's MAG1 and the CN's MAG2. As
LMA1 is the common anchor for the MN and the CN and maintains
location information for both nodes, no major race condition and
instability in updating the states for localized routing is
expected.
A22: The CN (CN2) connects to a different MAG (MAG2) and a different
LMA (LMA2) than the MN (MAG1, LMA1). The result of localized
routing should be the existence of routing information at MAG1 and
MAG2, which allows direct forwarding of packets between the MN's
MAG1 and the CN's MAG2. As the location information of the CN and
the MN is maintained at different LMAs, both LMAs need to be
involved in the procedure to set up localized routing. In the
case of a handover of the MN and/or the CN to a different MAG,
non-synchronized control of updating the states for localized
routing may result in race conditions, superfluous signaling, and
packet loss.
The following list summarizes general problems with setting up and
maintaining localized routing between an MN and a CN. In the context
of this problem statement, the MN and the CN are always assumed to be
registered at an LMA according to the PMIPv6 protocol [RFC5213].
o MNs do not participate in mobility management and hence cannot
perform binding registration at a CN on their own. Rather,
entities in the network infrastructure must take over the role of
MNs to set up and maintain a direct route. Accordingly, a
solution for localized routing in PMIPv6 must specify protocol
operation between relevant network components, such as between a
MAG and an LMA, to enable localized routing for data traffic
without traversing the MN's and the CN's LMA(s).
o In the case where the MN and the CN are both registered with
different LMAs according to the PMIPv6 protocol, relevant
information for the setup of a localized routing path, such as the
current MAG of the MN and the CN, is distributed between these
LMAs. This may complicate the setup and stable maintenance of
states enabling localized routing.
o In the case where localized routing between an MN and a CN has
been successfully set up and both nodes move and attach to a new
access router simultaneously, signaling the new location and
maintenance of states for localized routing at relevant routers
may run into a race condition situation. This can happen in the
case where coordination of signaling for localized routing and
provisioning of relevant state information is distributed between
different network entities, e.g., different LMAs. In such a case,
as a result of the MN's handover, updated information about the
MN's location may arrive at the CN's previous MAG, while the CN
has already moved to a new MAG. The same applies to the other
direction, where the system may update the MN's previous MAG about
the CN's new location, while the MN has moved to a new MAG in the
meantime. The protocol solution must deal with such exceptional
handover cases efficiently to avoid or resolve such problems.
3.3. IPv4 Considerations
According to [RFC5844], the basic configuration requirements for
supporting IPv4 in PMIPv6 are that LMAs and MAGs are both IPv4 and
IPv6 enabled. Also, LMAs and MAGs must have a globally unique IPv6
address configured, irrespective of enabled support for IPv6 routing
between these components. This requirement should also apply to
configuration requirements of localized routing.
Additional issues emerge when localized routing is considered for
PMIPv6 with IPv4 support. These can be classified into two general
groups: issues with localized routing between an MN's and a CN's IPv4
Home Addresses, and transport plane issues. The following
subsections analyze these two groups.
3.3.1. Localized Routing for Communication between IPv4 Home Addresses
In the case where an LMA and a MAG hold a registration to support
IPv4 Home Address mobility for an MN, the MAG and the LMA must
support appropriate encapsulation of IPv4 packets. To enable
localized routing, the MN's MAG must encapsulate and forward routing
path optimized packets to the CN's MAG and needs to ensure that the
chosen encapsulation mode is supported by the correspondent MAG.
Incompatibility in a selected encapsulation mode causes failure in
setting up a localized route.
When localized routing is used for IPv4 traffic, the conceptual data
structures on associated MAGs must be augmented with appropriate
parameters for forwarding localized traffic. MAGs may need to
maintain a routing state for each MN-CN-pair and make routing
decisions for uplink traffic based on the packet's complete IPv4
source and destination address. Hence, conceptual data structures to
handle states for localized routes need to comprise this address
tuple for unique identification.
As a known constraint, IPv4 addresses of two nodes that hold
addresses from a private address space may overlap. To uniquely
identify both nodes, the IPv4 address of the MN and the CN must not
overlap. To cope with overlapping address spaces, the localized
routing solution could use additional mechanisms to tag and uniquely
identify the MN and the CN.
3.3.2. IPv4 Transport Network Considerations
The transport network between the LMA and the MAG may be based on
IPv6 or IPv4. Deployments may ensure that the same transport
mechanism (i.e., IPv6 or IPv4) is used for operational consistency.
Similar to the encapsulation requirement stated in the previous
section, the IP version used for localized routing is also assumed,
by configuration, to be consistent across all MAGs within the
associated provider domain. The design of optional mechanisms for
negotiating the IP version to use as well as the encapsulation mode
to use are outside the scope of the NetExt WG's solution for
localized routing.
4. Functional Requirements for Localized Routing
Several tasks need to be performed by the network infrastructure
components before relevant information for such direct communication
is discovered and associated states for localized routing can be set
up. The following list summarizes some key functions that need to be
performed by the PMIPv6-enabled network infrastructure to substitute
mobile nodes in setting up a direct route.
o Detection of the possibility to perform localized routing. This
function includes looking at a data packet's source and
destination address.
o Initiation of a procedure that sets up a localized routing path.
o Discovery of stateful entities (i.e., the LMA(s) and/or the
MAG(s)) that maintain and can provide relevant information needed
to set up a localized routing path. Such information may include
the routable address of an LMA or MAG, where one or both mobile
nodes are connected to and registered with that LMA or MAG.
o Control in setting up and maintaining (e.g., during handover) the
localized routing path. Control is also needed to terminate the
use of a localized routing path and to delete associated states,
whereas a trigger for the termination may come from a non-PMIPv6-
related component.
o Enforcement of administrative policy rules to localized routing.
Such policies allow operators to have further control of the setup
of a localized route and enable the possibility to disallow
localized routing, for example, to ensure that traffic traverses
charging-related functions on the LMA. Explicit authorization of
localized routing is, for example, discussed in [PMIP6-LR]. As a
further example, mobile-node- and operator-specific policy rules
can be established on PMIPv6 components during PMIPv6
bootstrapping according to [RFC5779].
5. Roaming Considerations
Figure 2 shows PMIPv6 roaming cases where PMIPv6 components (e.g.,
LMAs, MAGs) tied by the MN and the CN may be distributed between
different provider domains (i.e., domain A, B, C) and the MN and/or
CN moves from one provider domain to another one. In order to
support localized routing when roaming occurs, it is required that
MAGs to which the MN and CN connect be within the same provider
domain, and each MAG has a security relationship with the
corresponding LMA, which maintains the registration of the MN or the
CN, respectively.
According to the roaming model as depicted in Figure 2, the MN's
PMIPv6 domain is characterized by its MAG (MAG1/MAG1') and its LMA
(LMA1), whereas the CN's PMIPv6 domain is characterized by the CN's
MAG (MAG2/MAG2') and its LMA (LMA2/LMA2'). A solution for localized
routing cannot take any assumption about whether or not the MN and CN
share the same PMIPv6 domain; hence, MAG1/MAG1' may not share a
security association with LMA2/LMA2', and MAG2/MAG2' may not share a
security association with LMA1, respectively.
It is not required that LMAs, which hold the registration for the MN
and the CN, respectively, be part of the same provider domain as the
MAGs where the MN and CN attach. When the MN's MAG and LMA belong to
different provider domains (A and C), localized routing is subject to
policy governing the service level agreements between these domains.
The same applies to the provider domains that provide the CN's MAG
and LMA. Based on the above requirements, four PMIPv6 roaming and
non-roaming cases can be taken into account.
o Case 1: The MN's MAG (MAG1), the CN's MAG (MAG2), the MN's LMA
(LMA1), and the CN's LMA (LMA2) are located in the same provider
domain A.
o Case 2: The MN's MAG (MAG1), the CN's MAG (MAG2), and the MN's LMA
(LMA1) are located in the same domain A, while the CN's LMA
(LMA2') is located in provider domain B.
o Case 3: The MN's MAG (MAG1') and the CN's MAG (MAG2') are located
in domain C, while the MN's LMA (LMA1) and the CN's LMA (LMA2) are
located in provider domain A.
o Case 4: The MN's MAG (MAG1') and the CN's MAG (MAG2') are located
in provider domain C, while the MN's LMA (LMA1) is located in
provider domain A and the CN's LMA (LMA2') is located in provider
domain B.
In these roaming cases, the MN can be allowed to roam within its
domain (e.g., the MN's home domain in which the MN's LMA is located)
or over different domains (e.g., the MN moves from its home domain to
a visited domain). During mobility, the CN and MN should remain
attached to MAGs of the same provider domain to maintain efficient
routing of traffic between their MAGs.
|
+-----+ +-----+ | +-----+
|LMA1 | |LMA2 | | |LMA2'|
+-----+ +-----+ | +-----+
|
|
|
|
+-----+ +-----+ |
|MAG1 | |MAG2 | |
+-----+ +-----+ |
|
|
Provider Domain A | Provider Domain B
------------------------------+-------------------------------
Provider Domain C
+-----+ +-----+
|MAG1'| |MAG2'|
+-----+ +-----+
Figure 2: PMIPv6 Roaming Cases Considered for Localized Routing
6. Security Considerations
A protocol solution for localized routing in a PMIPv6 network must
counter unauthorized change of a routing path. In particular, the
control plane for localized routing must preclude the blocking or
hijacking of mobile nodes' traffic by malicious or compromised
network components. A security solution must support suitable
mechanisms for authentication of control plane components of the
localized routing functional architecture for both roaming and
non-roaming scenarios. Any possibility for Internet hosts to
interfere with the localized routing procedure in a malicious manner
must be precluded.
Since network entities other than MNs and CNs perform signaling to
set up localized routing, the MIPv6 return routability test [RFC3775]
is not suitable to authenticate associated signaling messages in
PMIPv6. Solutions for localized routing in PMIPv6 need to mitigate,
or to provide sufficient defense against, possible security threats.
When PMIPv6 participants are administered within the same domain,
infrastructure-based authorization mechanisms, such as IPsec, may be
usable to protect signaling for localized routing.
Existing security associations according to [RFC5213] can be re-used
to protect signaling for localized routing on the interface between a
MAG and an LMA. In the case where a protocol solution for localized
routing in PMIPv6 relies on protocol operation between MAGs, means
for protection of signaling between these MAGs must be provided. The
same applies for signaling on a possible protocol interface between
two LMAs of the same domain.
7. Acknowledgments
Many aspects of the problem space for route optimization in PMIPv6
have been discussed in the context of a PMIPv6 Route Optimization
Design Goals document, which was submitted to the NetLMM WG in
November 2007. This group of contributors includes Sangjin Jeong,
Christian Vogt, Ryuji Wakikawa, Marco Liebsch, Behcet Sarikaya,
Shinta Sugimoto, Long Le, Alice Qinxia, and Jaehwoon Lee. Many
thanks to Rajeev Koodli for his comments about the structure and
scope of this problem statement for the NetExt WG.
This problem statement reflects the results of the discussion in the
NetExt group. Many thanks to Hidetoshi Yokota, Carlos Bernardos,
Ashutosh Dutta, Sri Gundavelli, Mohana Jeyatharan, Jouni Korhonen,
Glen Zorn, Dirk von Hugo, Frank Xia, Xiangsong Cui, and Basavaraj
Patil for their comments and support to improve the quality of the
problem statement.
8. References
8.1. Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC5213] Gundavelli, S., Ed., Leung, K., Devarapalli, V.,
Chowdhury, K., and B. Patil, "Proxy Mobile IPv6",
RFC 5213, August 2008.
[RFC5844] Wakikawa, R. and S. Gundavelli, "IPv4 Support for Proxy
Mobile IPv6", RFC 5844, May 2010.
8.2. Informative References
[PMIP6-LR] Zorn, G., Wu, Q., Liebsch, M., and J. Korhonen, "Diameter
Support for Proxy Mobile IPv6 Localized Routing", Work
in Progress, May 2011.
[RFC3775] Johnson, D., Perkins, C., and J. Arkko, "Mobility Support
in IPv6", RFC 3775, June 2004.
[RFC5779] Korhonen, J., Ed., Bournelle, J., Chowdhury, K., Muhanna,
A., and U. Meyer, "Diameter Proxy Mobile IPv6: Mobile
Access Gateway and Local Mobility Anchor Interaction with
Diameter Server", RFC 5779, February 2010.
Authors' Addresses
Marco Liebsch (editor)
NEC Laboratories Europe
NEC Europe Ltd.
Kurfuersten-Anlage 36
69115 Heidelberg
Germany
Phone: +49 6221 4342146
EMail: liebsch@neclab.eu
Sangjin Jeong
ETRI
218 Gajeongno, Yuseong
Daejeon 305-700
Korea
Phone: +82 42 860 1877
EMail: sjjeong@etri.re.kr
Qin Wu
Huawei Technologies Co., Ltd
101 Software Avenue, Yuhua District
Nanjing 210012
China
Phone: +86 25 56623633
EMail: sunseawq@huawei.com
ISO Plain Language
On this page
§1The Problem in One Paragraph
The IETF has specified Proxy Mobile IPv6 (RFC 5213) as the base protocol for network-based localized mobility management. The scope of the base protocol covers the setup and maintenance of a tunnel between a mobile node's Mobile Access Gateway and its selected Local Mobility Anchor. Data packets will always traverse the mobile node's gateway and its anchor, irrespective of the location of the mobile node's remote communication endpoint. Even though a node may be attached to the same gateway, or a different gateway, as its correspondent node within the same provider domain, packets associated with their communication will traverse both nodes' anchors, which can be located topologically far away from the gateways or even in a separate provider domain.
RFC 5213 addresses the need to enable local routing of traffic between two nodes attached to the same gateway, but does not specify the complete procedure to establish such localized routing state on the shared gateway.
The NetLMM Extensions working group has an objective to design a solution for localized routing in PMIPv6. That objective includes the specification of protocol messages and associated protocol operation between PMIPv6 components to support the setup of a direct routing path for data packets between the two gateways, while both hosts receive mobility service according to the PMIPv6 protocol. As a result of localized routing, packets are forwarded between the two gateways without traversing the anchors. In cases where one or both nodes hand over to a different gateway, the localized routing protocol maintains the localized routing path. Relevant protocol interfaces may include the interface between associated gateways, between a gateway and an anchor, and between anchors. The setup of localized routing with correspondent nodes not registered with a PMIPv6 network is out of scope of that solution and of this problem statement.
PMIPv6 keeps a mobile node's traffic anchored at a server in the operator's core network. Everything a node sends goes there and comes back, even when the node is talking to a machine on the same access network. This memo is about the case where the traffic could simply go from one access gateway to the other.
- Internet backbone
- LMA1
- MAG1
- MN
- CN1
- MAG1
- LMA2
- MAG2
- CN2
- MAG2
- LMA1
§2The Vocabulary
The document uses the terminology of RFC 5213 and adds four terms. A Mobile Node is a mobile node without IP mobility support, which is attached to a Mobile Access Gateway and registered with a Local Mobility Anchor according to the PMIPv6 specification. A Correspondent Node is the communication peer of such a mobile node, with or without IP mobility support.
Localized Routing is the result of signalling to set up routing states on the relevant network entities that allow forwarding of data packets between a mobile node and a correspondent node which are attached to gateways sharing the same provider domain, without the intervention of the mobile node's anchor and the correspondent node's anchor in data forwarding. Localized Routing States are the pieces of information for localized routing on the relevant forwarding entities on the optimised data path between the two nodes. Such information includes route entries and tunnel endpoints, and may include further information about the two nodes, such as their Mobile Node Identifier and their assigned Home Network Prefix. A Provider Domain is a network domain in which network components are administered by a single authority, for example the mobile operator.
Four terms matter here. The mobile node is the device that moves. The correspondent node is who it is talking to. Localized routing means the two talk directly instead of via the anchors. And a provider domain is the part of the network one organisation runs.
§3.1Why Operators Would Want Local Routing
Mobile IPv6 (RFC 3775) has built-in mechanisms for direct communication between a mobile node and a correspondent node. Mechanisms for route optimisation in Mobile IPv6 cannot be directly applied in PMIPv6: mobile nodes are not involved in mobility signalling and hence cannot perform signalling to set up localized routes. Instead, the solution for localized routing must consider functions in the network to find out whether a localized route is to be used and then control the setup and maintenance of localized routing states accordingly, without any assistance from the mobile node and the correspondent node.
Localized routing is understood in RFC 5213 as an optimisation of traffic between a mobile node and a correspondent node attached to an access link connected to the same gateway. In that case the gateway forwards traffic directly between the two, assuming it is enabled to support the feature through the setting of the EnableMAGLocalRouting flag, and the mobile node's anchor enforces the optimisation. RFC 5213 does not specify how an anchor can enforce the optimisation for such local communication, and maintaining local forwarding between the mobile node and a regular IPv6 correspondent node gets more complex when the mobile node hands over to a different gateway. This document focuses on use cases where both nodes are within a PMIPv6 network and served by an anchor in a domain of anchors.
With localized routing, operators have the possibility of offloading traffic from anchors and from the core network. Establishing a direct path between the two gateways is beneficial for two reasons. First, by limiting the communication to the access nodes, the data traffic traversing the gateway to anchor path can be reduced, which is significant given that the transport network between the access and the core is often the bottleneck in terms of costs and performance. Second, there may be performance benefits for data flows between the two nodes in terms of delay and packet loss, especially when both are attached to the same gateway and the anchor is topologically far away from it. Even when the nodes are attached to different gateways, there could be benefit in limiting the communication to the access network only. Offloading traffic from the anchor also improves the scalability of the anchor, since it is a bottleneck for traffic being forwarded by many gateways.
Two reasons to want the shortcut. The link between the access network and the operator's core is often the expensive, slow part, so keeping traffic off it saves money and reduces delay. And the anchor has to forward everything, so taking the easy traffic off it lets it handle the rest better.
§3.2The Use Cases and the Hard Parts
The problem statement focuses on local communication between PMIPv6 managed nodes which attach to gateways sharing the same provider domain. The use cases consider the existence of multiple anchors, and are named by the number of gateways and anchors involved: A11, A12, A21 and A22.
In A11, the mobile node and the correspondent node connect to the same gateway and are registered with the same anchor. The common gateway may forward packets between them directly without forwarding any packet to the anchor. RFC 5213 addresses this use case but does not specify the complete procedure to establish such localized routing state on the shared gateway. In A12, the two connect to the same gateway but are registered with different anchors. The common gateway may still forward packets directly, but potential problems exist in the case where the two anchors differ in their policy for controlling the gateway.
In A21, the correspondent node connects to a different gateway than the mobile node, but both are registered with the same anchor. The result of localized routing should be routing information at both gateways, allowing direct forwarding between them. Because that anchor is the common anchor for both nodes and maintains location information for both, no major race condition and instability in updating the states is expected. In A22, the correspondent node connects to a different gateway and a different anchor. Routing information is again needed at both gateways, but as the location information of the two nodes is maintained at different anchors, both anchors have to be involved in the procedure. In the case of a handover of one or both nodes to a different gateway, non-synchronised control of updating the states may result in race conditions, superfluous signalling and packet loss.
Three general problems follow. Mobile nodes do not participate in mobility management and hence cannot perform binding registration at a correspondent node on their own, so entities in the network infrastructure must take over that role, which means the solution must specify protocol operation between relevant network components such as between a gateway and an anchor. When the two nodes are registered with different anchors, the information needed to set up a localized path, such as the current gateway of each node, is distributed between those anchors, which may complicate the setup and stable maintenance of the states. And if both nodes move and attach to a new access router at the same time, signalling the new location may run into a race condition: as a result of the mobile node's handover, updated information about its location may arrive at the correspondent node's previous gateway while the correspondent node has already moved to a new gateway, and the same applies in the other direction. The protocol solution must deal with such exceptional handover cases efficiently.
The easy case is two devices behind one gateway and one anchor. The hard cases are when the anchors differ, because then the information needed to build the shortcut is in two places and has to be kept in step. The nastiest case is both devices moving at once, when each one's new location can arrive at the other's old gateway after the fact.
| Use case | Topology | What localized routing needs |
|---|---|---|
| A11 | Both nodes on the same gateway, registered with the same anchor | The shared gateway can forward directly; RFC 5213 addresses this but does not give the full procedure |
| A12 | Same gateway, different anchors | The shared gateway can still forward directly, but the two anchors may differ in their policy for controlling the gateway |
| A21 | Different gateways, same anchor | Routing state at both gateways; one anchor holds both locations, so no major race is expected |
| A22 | Different gateways, different anchors | Routing state at both gateways; both anchors must take part, so handovers can race |
§4What a Solution Has to Do
Several tasks need to be performed by the network infrastructure components before the relevant information for direct communication is discovered and the associated states can be set up. The document summarises the key functions that the PMIPv6-enabled infrastructure has to perform in order to substitute for the mobile nodes in setting up a direct route. These are: detection of the possibility to perform localized routing, which includes looking at a data packet's source and destination address; initiation of a procedure that sets up a localized routing path; discovery of the stateful entities, that is the anchors and/or gateways, that maintain and can provide the relevant information needed to set up the path, such as the routable address of an anchor or gateway where one or both mobile nodes are connected and registered; control in setting up and maintaining the path, for example during a handover, which is also needed to terminate the use of a localized path and delete the associated states, where a trigger for the termination may come from a component unrelated to PMIPv6; and enforcement of administrative policy rules, which allow operators to have further control over the setup of a localized route and enable the possibility of disallowing localized routing, for example to ensure that traffic traverses charging-related functions on the anchor. Explicit authorization of localized routing is discussed in a Diameter document, and mobile-node-specific and operator-specific policy rules can be established on PMIPv6 components during bootstrapping.
IPv4 support adds its own constraints. According to RFC 5844, the basic configuration requirements for supporting IPv4 in PMIPv6 are that anchors and gateways are both IPv4 and IPv6 enabled, and that both have a globally unique IPv6 address configured. Where an anchor and a gateway hold a registration supporting IPv4 home address mobility, they must support appropriate encapsulation of IPv4 packets, and the mobile node's gateway must ensure that the chosen encapsulation mode is supported by the correspondent gateway, since incompatibility causes failure in setting up a localized route. The routing state on the associated gateways must be keyed on the complete IPv4 source and destination address, and because IPv4 addresses from a private address space may overlap, the solution may need extra mechanisms to tag and identify the two nodes uniquely. The transport network between the anchor and the gateway may be based on IPv6 or IPv4, and the version used for localized routing is assumed by configuration to be consistent across all gateways within the associated provider domain.
Roaming adds a boundary. To support localized routing when roaming occurs, the gateways to which the two nodes connect must be within the same provider domain, and each gateway must have a security relationship with the corresponding anchor that holds the registration of the node. Four cases are considered, in which the two gateways and the two anchors are placed in the same or different provider domains. During mobility, the two nodes should remain attached to gateways of the same provider domain to maintain efficient routing of traffic between those gateways.
The infrastructure has to do five things: notice that a direct route is possible, start setting it up, find the equipment that holds the necessary records, keep the route working through handovers, and obey the operator's rules about when to allow it at all. With IPv4 in the mix, the two ends must also agree on how the packets are wrapped.
- Detect the possibilityLook at a data packet's source and destination address to see whether a direct route could be used.
- Start the setupInitiate the procedure that establishes the localized routing path.
- Find the right equipmentDiscover the anchors and gateways that hold the information needed, and their routable addresses.
- Control itSet the path up, maintain it during handovers, and tear it down when it should no longer be used.
- Apply policyEnforce administrative rules, including refusing to localize a route so that traffic still crosses the anchor's charging functions.
§6Security Considerations
A protocol solution for localized routing in a PMIPv6 network must counter unauthorized change of a routing path. In particular, the control plane for localized routing must preclude the blocking or hijacking of mobile nodes' traffic by malicious or compromised network components. A security solution must support suitable mechanisms for authentication of the control plane components of the localized routing functional architecture for both roaming and non-roaming scenarios. Any possibility for Internet hosts to interfere with the localized routing procedure in a malicious manner must be precluded.
Since network entities other than the mobile node and the correspondent node perform the signalling that sets up localized routing, the Mobile IPv6 return routability test (RFC 3775) is not suitable to authenticate associated signalling messages in PMIPv6. Solutions for localized routing in PMIPv6 need to mitigate, or provide sufficient defense against, possible security threats. When PMIPv6 participants are administered within the same domain, infrastructure-based authorization mechanisms such as IPsec may be usable to protect signalling for localized routing.
Existing security associations according to RFC 5213 can be re-used to protect signalling for localized routing on the interface between a gateway and an anchor. In the case where a protocol solution relies on protocol operation between gateways, means for the protection of signalling between those gateways must be provided. The same applies for signalling on a possible protocol interface between two anchors of the same domain.
Because the shortcut changes where traffic goes, an attacker who could push a routing change could divert a user's traffic. The network has to authenticate the parts that build the shortcut, and it can reuse the security associations PMIPv6 already sets up between a gateway and an anchor. A gateway to gateway link needs its own protection, because PMIPv6 does not create one today.