Network Working Group R. Koodli, Ed.
Request for Comments: 5568 Starent Networks
Obsoletes: 5268 July 2009
Category: Standards Track
Mobile IPv6 Fast Handovers
Mobile IPv6 enables a mobile node (MN) to maintain its connectivity
to the Internet when moving from one Access Router to another, a
process referred to as handover. During handover, there is a period
during which the mobile node is unable to send or receive packets
because of link-switching delay and IP protocol operations. This
"handover latency" resulting from standard Mobile IPv6 procedures
(namely, movement detection, new Care-of Address configuration, and
Binding Update) is often unacceptable to real-time traffic such as
Voice over IP (VoIP). Reducing the handover latency could be
beneficial to non-real-time, throughput-sensitive applications as
well. This document specifies a protocol to improve handover latency
due to Mobile IPv6 procedures. This document does not address
improving the link-switching latency.
This document updates the packet formats for the Handover Initiate
(HI) and Handover Acknowledge (HAck) messages to the Mobility Header
Status of This Memo
This document specifies an Internet standards track protocol for the
Internet community, and requests discussion and suggestions for
improvements. Please refer to the current edition of the "Internet
Official Protocol Standards" (STD 1) for the standardization state
and status of this protocol. Distribution of this memo is unlimited.
Copyright (c) 2009 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 in effect on the date of
publication of this document (http://trustee.ietf.org/license-info).
Please review these documents carefully, as they describe your rights
and restrictions with respect to this document.
Mobile IPv6 [RFC3775] describes the protocol operations for a mobile
node to maintain connectivity to the Internet during its handover
from one access router to another. These operations involve link-
layer procedures, movement detection, IP address configuration, and
location update. The combined handover latency is often sufficient
to affect real-time applications. Throughput-sensitive applications
can also benefit from reducing this latency. This document describes
a protocol to reduce the handover latency.
This specification addresses the following problems: how to allow a
mobile node to send packets as soon as it detects a new subnet link
and how to deliver packets to a mobile node as soon as its attachment
is detected by the new access router. The protocol defines IP
protocol messages necessary for its operation regardless of link
technology. It does this without depending on specific link-layer
features while allowing link-specific customizations. By definition,
this specification considers handovers that interwork with Mobile IP.
Once attached to its new access router, an MN engages in Mobile IP
operations including Return Routability [RFC3775]. There are no
special requirements for a mobile node to behave differently with
respect to its standard Mobile IP operations.
This specification is applicable when a mobile node has to perform
IP-layer operations as a result of handovers. This specification
does not address improving the link-switching latency. It does not
modify or optimize procedures related to signaling with the home
agent of a mobile node. Indeed, while targeted for Mobile IPv6, it
could be used with any mechanism that allows communication to
continue despite movements. Finally, this specification does not
address bulk movement of nodes using aggregate prefixes.
This document updates the protocol header format for the Handover
Initiate (HI) and Handover Acknowledge (HAck) messages defined in
[RFC5268]. Both the Proxy Mobile IPv6 (PMIPv6) protocol [RFC5213]
and the Mobile IPv6 protocol use Mobility Header (MH) as the type for
carrying signaling related to route updates. Even though the Fast
Handover protocol uses the Mobility Header for mobile node signaling
purposes, it has used ICMP for inter - access router communication.
Specifying Mobility Header for the HI and HAck messages enables
deployment of the protocol alongside PMIP6 and MIP6 protocols; the
reasons that led to this change are captured in Appendix B. Hence,
this document specifies the Mobility Header formats for HI and HAck
messages (Section 6.2.1) and the Mobility Header option format for
the IPv6 Address/Prefix option (Section 6.4.2), and deprecates the
use of ICMP for HI and HAck messages. Implementations of this
specification MUST NOT send ICMPv6 HI and HAck messages as defined in
[RFC5268]. If implementations of this specification receive ICMPv6
HI and HAck messages as defined in [RFC5268], they MAY interpret the
messages as defined in [RFC5268].
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 RFC 2119 [RFC2119].
The use of the term, "silently ignore" is not defined in RFC 2119.
However, the term is used in this document and can be similarly
The following terminology and abbreviations are used in this document
in addition to those defined in [RFC3775]. The reference handover
scenario is illustrated in Figure 1.
+-+ | Previous | <
| | ------------ | Access | ------- >-----\
+-+ | Router | < \
MN | (PAR) | \
| +--------------+ +---------------+
| ^ IP | Correspondent |
| | Network | Node |
V | +---------------+
v +--------------+ /
+-+ | New | < /
| | ------------ | Access | ------- >-----/
+-+ | Router | <
MN | (NAR) |
Figure 1: Reference Scenario for Handover
Mobile Node (MN): A Mobile IPv6 host.
Access Point (AP): A Layer 2 device connected to an IP subnet that
offers wireless connectivity to an MN. An Access Point Identifier
(AP-ID) refers the AP's L2 address. Sometimes, AP-ID is also
referred to as a Basic Service Set IDentifier (BSSID).
Access Router (AR): The MN's default router.
Previous Access Router (PAR): The MN's default router prior to its
New Access Router (NAR): The MN's anticipated default router
subsequent to its handover.
Previous CoA (PCoA): The MN's Care-of Address valid on PAR's
New CoA (NCoA): The MN's Care-of Address valid on NAR's subnet.
Handover: A process of terminating existing connectivity and
obtaining new IP connectivity.
Router Solicitation for Proxy Advertisement (RtSolPr): A message
from the MN to the PAR requesting information for a potential
Proxy Router Advertisement (PrRtAdv): A message from the PAR to
the MN that provides information about neighboring links
facilitating expedited movement detection. The message can also
act as a trigger for network-initiated handover.
[AP-ID, AR-Info] tuple: Contains an access router's L2 and IP
addresses, and prefix valid on the interface to which the Access
Point (identified by AP-ID) is attached. The triplet [Router's L2
address, Router's IP address, and Prefix] is called "AR-Info".
See also Section 5.3.
Neighborhood Discovery: The process of resolving neighborhood AP-
IDs to AR-Info.
Assigned Addressing: A particular type of NCoA configuration in
which the NAR assigns an IPv6 address for the MN. The method by
which the NAR manages its address pool is not specified in this
Fast Binding Update (FBU): A message from the MN instructing its
PAR to redirect its traffic (toward NAR).
Fast Binding Acknowledgment (FBack): A message from the PAR in
response to an FBU.
Predictive Fast Handover: The fast handover in which an MN is able
to send an FBU when it is attached to the PAR, which then
establishes forwarding for its traffic (even before the MN
attaches to the NAR).
Reactive Fast Handover: The fast handover in which an MN is able
to send the FBU only after attaching to the NAR.
Unsolicited Neighbor Advertisement (UNA): The message in [RFC4861]
with 'O' bit cleared.
Fast Neighbor Advertisement (FNA): This message from RFC 4068
[RFC4068] is deprecated. The UNA message above is the preferred
message in this specification.
Handover Initiate (HI): A message from the PAR to the NAR
regarding an MN's handover.
Handover Acknowledge (HAck): A message from the NAR to the PAR as
a response to HI.
3. Protocol Overview
3.1. Addressing the Handover Latency
The ability to immediately send packets from a new subnet link
depends on the "IP connectivity" latency, which in turn depends on
the movement detection latency and the new CoA configuration latency.
Once an MN is IP-capable on the new subnet link, it can send a
Binding Update to its Home Agent and one or more correspondents.
Once its correspondents process the Binding Update successfully,
which typically involves the Return Routability procedure, the MN can
receive packets at the new CoA. So, the ability to receive packets
from correspondents directly at its new CoA depends on the Binding
Update latency as well as the IP connectivity latency.
The protocol enables an MN to quickly detect that it has moved to a
new subnet by providing the new access point and the associated
subnet prefix information when the MN is still connected to its
current subnet (i.e., PAR in Figure 1). For instance, an MN may
discover available access points using link-layer-specific mechanisms
(e.g., a "scan" in a Wireless Local Area Network (WLAN)) and then
request subnet information corresponding to one or more of those
discovered access points. The MN may do this after performing router
discovery or at any time while connected to its current router. The
result of resolving an identifier associated with an access point is
an [AP-ID, AR-Info] tuple, which an MN can use in readily detecting
movement. When attachment to an access point with AP-ID takes place,
the MN knows the corresponding new router's coordinates including its
prefix, IP address, and L2 address. The "Router Solicitation for
Proxy Advertisement (RtSolPr)" and "Proxy Router Advertisement
(PrRtAdv)" messages in Section 6.1 are used for aiding movement
Through the RtSolPr and PrRtAdv messages, the MN also formulates a
prospective new CoA (NCoA) when it is still present on the PAR's
link. Hence, the latency due to new prefix discovery subsequent to
handover is eliminated. Furthermore, this prospective address can be
used immediately after attaching to the new subnet link (i.e., NAR's
link) when the MN has received a "Fast Binding Acknowledgment
(FBack)" (see Section 6.2.3) message prior to its movement. In the
event it moves without receiving an FBack, the MN can still start
using NCoA after announcing its attachment through an unsolicited
Neighbor Advertisement message (with the 'O' bit set to zero)
[RFC4861]; NAR responds to this UNA message in case it wishes to
provide a different IP address to use. In this way, NCoA
configuration latency is reduced.
The information provided in the PrRtAdv message can be used even when
DHCP [RFC3315] is used to configure an NCoA on the NAR's link. In
this case, the protocol supports forwarding using PCoA, and the MN
performs DHCP once it attaches to the NAR's link. The MN still
formulates an NCoA for FBU processing; however, it MUST NOT send data
packets using the NCoA in the FBU.
In order to reduce the Binding Update latency, the protocol specifies
a binding between the Previous CoA (PCoA) and NCoA. An MN sends a
"Fast Binding Update" (see Section 6.2.2) message to its Previous
Access Router to establish this tunnel. When feasible, the MN SHOULD
send an FBU from the PAR's link. Otherwise, the MN should send the
FBU immediately after detecting attachment to the NAR. An FBU
message MUST contain the Binding Authorization Data for FMIPv6 (BADF)
option (see Section 6.4.5) in order to ensure that only a legitimate
MN that owns the PCoA is able to establish a binding. Subsequent
sections describe the protocol mechanics. In any case, the result is
that the PAR begins tunneling packets arriving for PCoA to NCoA.
Such a tunnel remains active until the MN completes the Binding
Update with its correspondents. In the opposite direction, the MN
SHOULD reverse tunnel packets to the PAR, again until it completes
the Binding Update. And, PAR MUST forward the inner packet in the
tunnel to its destination (i.e., to the MN's correspondent). Such a
reverse tunnel ensures that packets containing a PCoA as a source IP
address are not dropped due to ingress filtering. Even though the MN
is IP-capable on the new link, it cannot use the NCoA directly with
its correspondents without the correspondents first establishing a
binding cache entry (for the NCoA). Forwarding support for the PCoA
is provided through a reverse tunnel between the MN and the PAR.
Setting up a tunnel alone does not ensure that the MN receives
packets as soon as it is attached to a new subnet link, unless the
NAR can detect the MN's presence. A neighbor discovery operation
involving a neighbor's address resolution (i.e., Neighbor
Solicitation and Neighbor Advertisement) typically results in
considerable delay, sometimes lasting multiple seconds. For
instance, when arriving packets trigger the NAR to send Neighbor
Solicitation before the MN attaches, subsequent retransmissions of
address resolution are separated by a default period of one second
each. In order to circumvent this delay, an MN announces its
attachment immediately with an UNA message that allows the NAR to
forward packets to the MN right away. Through tunnel establishment
for PCoA and fast advertisement, the protocol provides expedited
forwarding of packets to the MN.
The protocol also provides the following important functionalities.
The access routers can exchange messages to confirm that a proposed
NCoA is acceptable. For instance, when an MN sends an FBU from the
PAR's link, FBack can be delivered after the NAR considers the NCoA
acceptable for use. This is especially useful when addresses are
assigned by the access router. The NAR can also rely on its trust
relationship with the PAR before providing forwarding support for the
MN. That is, it may create a forwarding entry for the NCoA, subject
to "approval" from the PAR, which it trusts. In addition, buffering
for handover traffic at the NAR may be desirable. Even though the
Neighbor Discovery protocol provides a small buffer (typically one or
two packets) for packets awaiting address resolution, this buffer may
be inadequate for traffic, such as VoIP, already in progress. The
routers may also wish to maintain a separate buffer for servicing the
handover traffic. Finally, the access routers could transfer
network-resident contexts, such as access control, Quality of Service
(QoS), and header compression, in conjunction with handover (although
the context transfer process itself is not specified in this
document). For all these operations, the protocol provides "Handover
Initiate (HI)" and "Handover Acknowledge (HAck)" messages (see
Section 6.2.1). Both of these messages SHOULD be used. The access
routers MUST have the necessary security association established by
means outside the scope of this document.
3.2. Protocol Operation
The protocol begins when an MN sends an RtSolPr message to its access
router to resolve one or more Access Point Identifiers to subnet-
specific information. In response, the access router (e.g., PAR in
Figure 1) sends a PrRtAdv message containing one or more [AP-ID,
AR-Info] tuples. The MN may send an RtSolPr at any convenient time,
for instance as a response to some link-specific event (a "trigger")
or simply after performing router discovery. However, the
expectation is that prior to sending an RtSolPr, the MN will have
discovered the available APs by link-specific methods. The RtSolPr
and PrRtAdv messages do not establish any state at the access router;
their packet formats are defined in Section 6.1.
With the information provided in the PrRtAdv message, the MN
formulates a prospective NCoA and sends an FBU message to the PAR.
The purpose of the FBU is to authorize the PAR to bind the PCoA to
the NCoA, so that arriving packets can be tunneled to the new
location of the MN. The FBU should be sent from the PAR's link
whenever feasible. For instance, an internal link-specific trigger
could enable FBU transmission from the previous link.
When it is not feasible, the FBU is sent from the new link.
The format and semantics of FBU processing are specified in
Section 6.2.2. The FBU message MUST contain the BADF option (see
Section 6.4.5) to secure the message.
Depending on whether an FBack is received on the previous link (which
clearly depends on whether the FBU was sent in the first place),
there are two modes of operation.
1. The MN receives FBack on the previous link. This means that
packet tunneling is already in progress by the time the MN
handovers to the NAR. The MN SHOULD send the UNA immediately
after attaching to the NAR, so that arriving as well as buffered
packets can be forwarded to the MN right away. Before sending
FBack to the MN, the PAR can determine whether the NCoA is
acceptable to the NAR through the exchange of HI and HAck
messages. When Assigned Addressing (i.e., addresses are assigned
by the router) is used, the proposed NCoA in the FBU is carried
in an HI message (from PAR to NAR), and NAR MAY assign the
proposed NCoA. Such an assigned NCoA MUST be returned in HAck
(from NAR to PAR), and PAR MUST in turn provide the assigned NCoA
in FBack. If there is an assigned NCoA returned in FBack, the MN
MUST use the assigned address (and not the proposed address in
FBU) upon attaching to NAR.
2. The MN does not receive the FBack on the previous link because
the MN has not sent the FBU or the MN has left the link after
sending the FBU (which itself may be lost), but before receiving
an FBack. Without receiving an FBack in the latter case, the MN
cannot ascertain whether the PAR has processed the FBU
successfully. Hence, the MN (re)sends the FBU message to the PAR
immediately after sending the UNA message. If the NAR chooses to
supply a different IP address to use than the NCoA, it MAY send a
Router Advertisement with the "Neighbor Advertisement Acknowledge
(NAACK)" option in which it includes an alternate IP address for
the MN to use. Detailed UNA processing rules are specified in
The scenario in which an MN sends an FBU and receives an FBack on
PAR's link is illustrated in Figure 2. For convenience, this
scenario is characterized as the "predictive" mode of operation. The
scenario in which the MN sends an FBU from the NAR's link is
illustrated in Figure 3. For convenience, this scenario is
characterized as the "reactive" mode of operation. Note that the
reactive mode also includes the case in which an FBU has been sent
from the PAR's link, but an FBack has not yet been received. The
figure is intended to illustrate that the FBU is forwarded through
the NAR, but it is processed only by the PAR.
MN PAR NAR
| | |
| | |
| <--FBack---|--FBack---> |
| | |
disconnect forward |
| packets ===============>|
| | |
| | |
connect | |
| | |
|<=================================== deliver packets
Figure 2: Predictive Fast Handover
MN PAR NAR
| | |
| | |
disconnect | |
| | |
| | |
connect | |
| |(HI/HAck if necessary)|
| forward |
| packets(including FBAck)=====>|
| | |
|<=================================== deliver packets
Figure 3: Reactive Fast Handover
Finally, the PrRtAdv message may be sent unsolicited, i.e., without
the MN first sending an RtSolPr. This mode is described in
3.3. Protocol Operation during Network-Initiated Handover
In some wireless technologies, the handover control may reside in the
network even though the decision to undergo handover may be mutually
arrived at between the MN and the network. In such networks, the PAR
can send an unsolicited PrRtAdv containing the link-layer address, IP
address, and subnet prefix of the NAR when the network decides that a
handover is imminent. The MN MUST process this PrRtAdv to configure
a new Care-of Address on the new subnet, and MUST send an FBU to the
PAR prior to switching to the new link. After transmitting PrRtAdv,
the PAR MUST continue to forward packets to the MN on its current
link until the FBU is received. The rest of the operation is the
same as that described in Section 3.2.
The unsolicited PrRtAdv also allows the network to inform the MN
about geographically adjacent subnets without the MN having to
explicitly request that information. This can reduce the amount of
wireless traffic required for the MN to obtain a neighborhood
topology map of links and subnets. Such usage of PrRtAdv is
decoupled from the actual handover; see Section 6.1.2.
4. Protocol Details
All descriptions refer to Figure 1.
After discovering one or more nearby access points, the MN sends
RtSolPr to the PAR in order to resolve access point identifiers to
subnet router information. A convenient time to do this is after
performing router discovery. However, the MN can send RtSolPr at any
time, e.g., when one or more new access points are discovered. The
MN can also send RtSolPr more than once during its attachment to PAR.
The trigger for sending RtSolPr can originate from a link-specific
event, such as the promise of a better signal strength from another
access point coupled with fading signal quality with the current
access point. Such events, often broadly referred to as "L2
triggers", are outside the scope of this document. Nevertheless,
they serve as events that invoke this protocol. For instance, when a
"link up" indication is obtained on the new link, protocol messages
(e.g., UNA) can be transmitted immediately. Implementations SHOULD
make use of such triggers whenever available.
The RtSolPr message contains one or more AP-IDs. A wildcard requests
all available tuples.
As a response to RtSolPr, the PAR sends a PrRtAdv message that
indicates one of the following possible conditions.
1. If the PAR does not have an entry corresponding to the new access
point, it MUST respond indicating that the new access point is
unknown. The MN MUST stop fast handover protocol operations on
the current link. The MN MAY send an FBU from its new link.
2. If the new access point is connected to the PAR's current
interface (to which the MN is attached), the PAR MUST respond
with a Code value indicating that the new access point is
connected to the current interface, but not send any prefix
information. This scenario could arise, for example, when
several wireless access points are bridged into a wired network.
No further protocol action is necessary.
3. If the new access point is known and the PAR has information
about it, then the PAR MUST respond indicating that the new
access point is known and supply the [AP-ID, AR-Info] tuple. If
the new access point is known, but does not support fast
handover, the PAR MUST indicate this with Code 3 (see
4. If a wildcard is supplied as an identifier for the new access
point, the PAR SHOULD supply neighborhood [AP-ID, AR-Info] tuples
that are subject to path MTU restrictions (i.e., provide any 'n'
tuples without exceeding the link MTU).
When further protocol action is necessary, some implementations MAY
choose to begin buffering copies of incoming packets at the PAR. If
such First In First Out (FIFO) buffering is used, the PAR MUST
continue forwarding the packets to the PCoA (i.e., buffer and
forward). While the protocol does not forbid such an implementation
support, care must be taken to ensure that the PAR continues
forwarding packets to the PCoA (i.e., uses a buffer and forward
approach). The PAR SHOULD stop buffering once it begins forwarding
packets to the NCoA.
The method by which access routers exchange information about their
neighbors and thereby allow construction of Proxy Router
Advertisements with information about neighboring subnets is outside
the scope of this document.
The RtSolPr and PrRtAdv messages MUST be implemented by an MN and an
access router that supports fast handovers. However, when the
parameters necessary for the MN to send packets immediately upon
attaching to the NAR are supplied by the link-layer handover
mechanism itself, use of the above messages is optional on such
After a PrRtAdv message is processed, the MN sends an FBU at a time
determined by link-specific events, and includes the proposed NCoA.
The MN SHOULD send the FBU from the PAR's link whenever
"anticipation" of handover is feasible. When anticipation is not
feasible or when it has not received an FBack, the MN sends an FBU
immediately after attaching to NAR's link. In response to the FBU,
the PAR establishes a binding between the PCoA ("Home Address") and
the NCoA, and sends the FBack to the MN. Prior to establishing this
binding, the PAR SHOULD send an HI message to the NAR, and receive a
HAck in response. In order to determine the NAR's address for the HI
message, the PAR can perform the longest prefix match of NCoA (in
FBU) with the prefix list of neighboring access routers. When the
source IP address of the FBU is the PCoA, i.e., the FBU is sent from
the PAR's link, the HI message MUST have a Code value set to 0; see
Section 184.108.40.206. When the source IP address of the FBU is not PCoA,
i.e., the FBU is sent from the NAR's link, the HI message MUST have a
Code value of 1; see Section 220.127.116.11.
The HI message contains the PCoA, link-layer address and the NCoA of
the MN. In response to processing an HI message with Code 0, the
1. determines whether the NCoA supplied in the HI message is unique
before beginning to defend it. It sends a Duplicate Address
Detection (DAD) probe [RFC4862] for NCoA to verify uniqueness.
However, in deployments where the probability of address
collisions is considered extremely low (and hence not an issue),
the parameter DupAddrDetectTransmits (see [RFC4862]) is set to
zero on the NAR, allowing it to avoid performing DAD on the NCoA.
The NAR similarly sets DupAddrDetectTransmits to zero in other
deployments where DAD is not a concern. Once the NCoA is
determined to be unique, the NAR starts proxying [RFC4861] the
address for PROXY_ND_LIFETIME during which the MN is expected to
connect to the NAR. In case there is already an NCoA present in
its data structure (for instance, it has already processed an HI
message earlier), the NAR MAY verify if the LLA is the same as
its own or that of the MN itself. If so, the NAR MAY allow the
use of the NCoA.
2. allocates the NCoA for the MN when Assigned Addressing is used,
creates a proxy neighbor cache entry, and begins defending it.
The NAR MAY allocate the NCoA proposed in HI.
3. MAY create a host route entry for the PCoA (on the interface to
which the MN is attaching) in case the NCoA cannot be accepted or
assigned. This host route entry SHOULD be implemented such that
until the MN's presence is detected, either through explicit
announcement by the MN or by other means, arriving packets do not
invoke neighbor discovery. The NAR SHOULD also set up a reverse
tunnel to the PAR in this case.
4. provides the status of the handover request in the Handover
Acknowledge (HAck) message to the PAR.
When the Code value in HI is 1, the NAR MUST skip the above
operations. Sending an HI message with Code 1 allows the NAR to
validate the neighbor cache entry it creates for the MN during UNA
processing. That is, the NAR can make use of the knowledge that its
trusted peer (i.e., the PAR) has a trust relationship with the MN.
If HAck contains an assigned NCoA, the FBack MUST include it, and the
MN MUST use the address provided in the FBack. The PAR MAY send the
FBack to the previous link as well to facilitate faster reception in
the event that the MN is still present. The result of the FBU and
FBack processing is that the PAR begins tunneling the MN's packets to
the NCoA. If the MN does not receive an FBack message even after
retransmitting the FBU for FBU_RETRIES, it must assume that fast
handover support is not available and stop the protocol operation.
As soon as the MN establishes link connectivity with the NAR, it:
1. sends an UNA message (see Section 6.3). If the MN has not
received an FBack by the time UNA is being sent, it SHOULD send
an FBU message following the UNA message.
2. joins the all-nodes multicast group and the solicited-node
multicast group corresponding to the NCoA.
3. starts a DAD probe for NCoA; see [RFC4862].
When a NAR receives an UNA message, it:
1. deletes its proxy neighbor cache entry, if it exists, updates the
state to STALE [RFC4861], and forwards arriving and buffered
2. updates an entry in INCOMPLETE state [RFC4861], if it exists, to
STALE and forwards arriving and buffered packets. This would be
the case if NAR had previously sent a Neighbor Solicitation that
went unanswered perhaps because the MN had not yet attached to
The buffer for handover traffic should be linked to this UNA
processing. The exact mechanism is implementation dependent.
The NAR may choose to provide a different IP address other than the
NCoA. This is possible if it is proxying the NCoA. In such a case,
1. MAY send a Router Advertisement with the NAACK option in which it
includes an alternate IP address for use. This message MUST be
sent to the source IP address present in UNA using the same Layer
2 address present in UNA.
If the MN receives an IP address in the NAACK option, it MUST use it
and send an FBU using the new CoA. As a special case, the address
supplied in NAACK could be the PCoA itself, in which case the MN MUST
NOT send any more FBUs. The Status codes for the NAACK option are
specified in Section 6.4.6.
Once the MN has confirmed its NCoA (either through DAD or when
provided for by the NAR), it SHOULD send a Neighbor Advertisement
message with the 'O' bit set, to the all-nodes multicast address.
This message allows the MN's neighbors to update their neighbor cache
For data forwarding, the PAR tunnels packets using its global IP
address valid on the interface to which the MN was attached. The MN
reverse tunnels its packets to the same global address of PAR. The
tunnel end-point addresses must be configured accordingly. When the
PAR receives a reverse-tunneled packet, it must verify if a secure
binding exists for the MN identified by the PCoA in the tunneled
packet, before forwarding the packet.
5. Other Considerations
5.1. Handover Capability Exchange
The MN expects a PrRtAdv in response to its RtSolPr message. If the
MN does not receive a PrRtAdv message even after RTSOLPR_RETRIES, it
must assume that the PAR does not support the fast handover protocol
and stop sending any more RtSolPr messages.
Even if an MN's current access router is capable of providing fast
handover support, the new access router to which the MN attaches may
be incapable of fast handover. This is indicated to the MN during
"runtime", through the PrRtAdv message with Code 3 (see
5.2. Determining New Care-of Address
Typically, the MN formulates its prospective NCoA using the
information provided in a PrRtAdv message and sends the FBU. The PAR
MUST use the NCoA present in the FBU in its HI message. The NAR MUST
verify if the NCoA present in HI is already in use. In any case, the
NAR MUST respond to HI using a HAck, in which it may include another
NCoA to use, especially when assigned address configuration is used.
If there is a CoA present in the HAck, the PAR MUST include it in the
FBack message. However, the MN itself does not have to wait on the
PAR's link for this exchange to take place. It can handover any time
after sending the FBU message; sometimes it may be forced to handover
without sending the FBU. In any case, it can still confirm using the
NCoA from the NAR's link by sending the UNA message.
If a PrRtAdv message carries an NCoA, the MN MUST use it as its
When DHCP is used, the protocol supports forwarding for the PCoA
only. In this case, the MN MUST perform DHCP operations once it
attaches to the NAR even though it formulates an NCoA for
transmitting the FBU. This is indicated in the PrRtAdv message with
5.3. Prefix Management
As defined in Section 2, the Prefix part of "AR-Info" is the prefix
valid on the interface to which the AP is attached. This document
does not specify how this Prefix is managed, its length, or its
assignment policies. The protocol operation specified in this
document works regardless of these considerations. Often, but not
necessarily always, this Prefix may be the aggregate prefix (such as
/48) valid on the interface. In some deployments, each MN may have
its own per-mobile prefix (such as a /64) used for generating the
NCoA. Some point-to-point links may use such a deployment.
When per-mobile prefix assignment is used, the "AR-Info" advertised
in PrRtAdv still includes the (aggregate) prefix valid on the
interface to which the target AP is attached, unless the access
routers communicate with each other (using HI and HAck messages) to
manage the per-mobile prefix. The MN still formulates an NCoA using
the aggregate prefix. However, an alternate NCoA based on the per-
mobile prefix is returned by NAR in the HAck message. This alternate
NCoA is provided to the MN in either the FBack message or in the
5.4. Packet Loss
Handover involves link switching, which may not be exactly
coordinated with fast handover signaling. Furthermore, the arrival
pattern of packets is dependent on many factors, including
application characteristics, network queuing behaviors, etc. Hence,
packets may arrive at the NAR before the MN is able to establish its
link there. These packets will be lost unless they are buffered by
the NAR. Similarly, if the MN attaches to the NAR and then sends an
FBU message, packets arriving at the PAR until the FBU is processed
will be lost unless they are buffered. This protocol provides an
option to indicate a request for buffering at the NAR in the HI
message. When the PAR requests this feature (for the MN), it SHOULD
also provide its own support for buffering.
Whereas buffering can enable a smooth handover, the buffer size and
the rate at which buffered packets are eventually forwarded are
important considerations when providing buffering support. There are
a number of aspects to consider:
o Some applications transmit less data over a given period of data
than others, and this implies different buffering requirements.
For instance, Voice over IP typically needs smaller buffers
compared to high-resolution streaming video, as the latter has
larger packet sizes and higher arrival rates.
o When the mobile node appears on the new link, having the buffering
router send a large number of packets in quick succession may
overtax the resources of the router, the mobile node itself, or
the path between these two.
In particular, transmitting a large amount of buffered packets in
succession can congest the path between the buffering router and
the mobile node. Furthermore, nodes (such as a base station) on
the path between the buffering router and the mobile node may drop
such packets. If a base station buffers too many such packets,
they may contribute to additional jitter for packets arriving
behind them, which is undesirable for real-time communication.
o Since routers are not involved in end-to-end communication, they
have no knowledge of transport conditions.
o The wireless connectivity of the mobile node may vary over time.
It may achieve a smaller or higher bandwidth on the new link,
signal strength may be weak at the time it just enters the area of
this access point, and so on.
As a result, it is difficult to design an algorithm that would
transmit buffered packets at appropriate spacing under all scenarios.
The purpose of fast handovers is to avoid packet loss. Yet, draining
buffered packets too fast can, by itself, cause loss of the packets,
as well as blocking or loss of following packets meant for the mobile
This specification does not restrict implementations from providing
specialized buffering support for any specific situation. However,
attention must be paid to the rate at which buffered packets are
forwarded to the MN once attachment is complete. Routers
implementing this specification MUST implement at least the default
algorithm, which is based on the original arrival rates of the
buffered packets. A maximum of 5 packets MAY be sent one after
another, but all subsequent packets SHOULD use a sending rate that is
determined by metering the rate at which packets have entered the
buffer, potentially using smoothing techniques such as recent
activity over a sliding time window and weighted averages [RFC3290].
It should be noted, however, that this default algorithm is crude and
may not be suitable for all situations. Future revisions of this
specification may provide additional algorithms, once enough
experience of the various conditions in deployed networks is
5.5. DAD Handling
Duplicate Address Detection (DAD) was defined in [RFC4862] to avoid
address duplication on links when stateless address auto-
configuration is used. The use of DAD to verify the uniqueness of an
IPv6 address configured through stateless auto-configuration adds
delays to a handover. The probability of an interface identifier
duplication on the same subnet is very low; however, it cannot be
ignored. Hence, the protocol specified in this document SHOULD only
be used in deployments where the probability of such address
collisions is extremely low or it is not a concern (because of the
address management procedure deployed). The protocol requires the
NAR to send a DAD probe before it starts defending the NCoA.
However, this DAD delay can be turned off by setting
DupAddrDetectTransmits to zero on the NAR ([RFC4862]).
This document specifies messages that can be used to provide
duplicate-free addresses, but the document does not specify how to
create or manage such duplicate-free addresses. In some cases, the
NAR may already have the knowledge required to assess whether or not
the MN's address is a duplicate before the MN moves to the new
subnet. For example, in some deployments, the NAR may maintain a
pool of duplicate-free addresses in a list for handover purposes. In
such cases, the NAR can provide this disposition in the HAck message
(see Section 18.104.22.168) or in the NAACK option (see Section 6.4.6).
5.6. Fast or Erroneous Movement
Although this specification is for fast handover, the protocol is
limited in terms of how fast an MN can move. A special case of fast
movement is ping-pong, where an MN moves between the same two access
points rapidly. Another instance of the same problem is erroneous
movement, i.e., the MN receives information prior to a handover that
it is moving to a new access point, but it either moves to a
different one or it aborts movement altogether. All of the above
behaviors are usually the result of link-layer idiosyncrasies and
thus are often resolved at the link layer itself.
IP layer mobility, however, introduces its own limits. IP-layer
handovers should occur at a rate suitable for the MN to update the
binding of, at least, its Home Agent and preferably that of every
correspondent node (CN) with which it is in communication. An MN
that moves faster than necessary for this signaling to complete
(which may be of the order of a few seconds) may start losing
packets. The signaling cost over the air interface and in the
network may increase significantly, especially in the case of rapid
movement between several access routers. To avoid the signaling
overhead, the following measures are suggested.
An MN returning to the PAR before updating the necessary bindings
when present on the NAR MUST send a Fast Binding Update with the Home
Address equal to the MN's PCoA and a lifetime of zero to the PAR.
The MN should have a security association with the PAR since it
performed a fast handover to the NAR. The PAR, upon receiving this
Fast Binding Update, will check its set of outgoing (temporary fast
handover) tunnels. If it finds a match, it SHOULD terminate that
tunnel; i.e., start delivering packets directly to the node instead.
In order for the PAR to process such an FBU, the lifetime of the
security association has to be at least that of the tunnel itself.
Temporary tunnels for the purposes of fast handovers should use short
lifetimes (of the order of tens of seconds). The lifetime of such
tunnels should be enough to allow an MN to update all its active
bindings. The default lifetime of the tunnel should be the same as
the lifetime value in the FBU message.
The effect of erroneous movement is typically limited to the loss of
packets since routing can change and the PAR may forward packets
toward another router before the MN actually connects to that router.
If the MN discovers itself on an unanticipated access router, it
SHOULD send a new Fast Binding Update to the PAR. This FBU
supersedes the existing binding at the PAR, and the packets will be
redirected to the newly confirmed location of the MN.