Request for Comments: 4260 Lucent Technologies
Category: Informational November 2005
Mobile IPv6 Fast Handovers for 802.11 Networks
Status of This Memo
This memo provides information for the Internet community. It does
not specify an Internet standard of any kind. Distribution of this
memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2005).
Abstract
This document describes how a Mobile IPv6 Fast Handover could be
implemented on link layers conforming to the 802.11 suite of
specifications.
Table of Contents
1. Introduction ....................................................2
1.1. Conventions Used in This Document ..........................2
2. Terminology .....................................................2
3. Deployment Architectures for Mobile IPv6 on 802.11 ..............3
4. 802.11 Handovers in Detail ......................................5
5. FMIPv6 Message Exchanges ........................................7
6. Beacon Scanning and NAR Discovery ...............................8
7. Scenarios .......................................................9
7.1. Scenario 1abcdef23456g .....................................9
7.2. Scenario ab123456cdefg ....................................10
7.3. Scenario 123456abcdefg ....................................10
8. Security Considerations ........................................10
9. Conclusions ....................................................12
10. References ....................................................13
10.1. Normative References .....................................13
10.2. Informative References ...................................13
11. Acknowledgements ..............................................13
1. Introduction
The Mobile IPv6 Fast Handover protocol [2] has been proposed as a way
to minimize the interruption in service experienced by a Mobile IPv6
node as it changes its point of attachment to the Internet. Without
such a mechanism, a mobile node cannot send or receive packets from
the time that it disconnects from one point of attachment in one
subnet to the time it registers a new care-of address from the new
point of attachment in a new subnet. Such an interruption would be
unacceptable for real-time services such as Voice-over-IP.
The basic idea behind a Mobile IPv6 fast handover is to leverage
information from the link-layer technology to either predict or
rapidly respond to a handover event. This allows IP connectivity to
be restored at the new point of attachment sooner than would
otherwise be possible. By tunneling data between the old and new
access routers, it is possible to provide IP connectivity in advance
of actual Mobile IP registration with the home agent or correspondent
node. This allows real-time services to be reestablished without
waiting for such Mobile IP registration to complete. Because Mobile
IP registration involves time-consuming Internet round-trips, the
Mobile IPv6 fast handover can provide for a smaller interruption in
real-time services than an ordinary Mobile IP handover.
The particular link-layer information available, as well as the
timing of its availability (before, during, or after a handover has
occurred), differs according to the particular link-layer technology
in use. This document gives a set of deployment examples for Mobile
IPv6 Fast Handovers on 802.11 networks. We begin with a brief
overview of relevant aspects of basic 802.11 [3]. We examine how and
when handover information might become available to the IP layers
that implement Fast Handover, both in the network infrastructure and
on the mobile node. Finally, we trace the protocol steps for Mobile
IPv6 Fast Handover in this environment.
1.1. 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 RFC 2119 [1].
2. Terminology
This document borrows all of the terminology from Mobile IPv6 Fast
Handovers [2], with the following additional terms from the 802.11
specification [3] (some definitions slightly modified for clarity):
Access Point (AP): Any entity that has station functionality and
provides access to the distribution services, via the
wireless medium (WM) for associated stations.
Association: The service used to establish access point/station
(AP/STA) mapping and enable STA access to the
Distribution System.
Basic Service Set (BSS): A set of stations controlled by a single
coordination function, where the coordination function
may be centralized (e.g., in a single AP) or
distributed (e.g., for an ad hoc network). The BSS
can be thought of as the coverage area of a single AP.
Distribution System (DS): A system used to interconnect a set of
basic service sets (BSSs) and integrated local area
networks (LANs) to create an extended service set
(ESS).
Extended Service Set (ESS): A set of one or more interconnected basic
service sets (BSSs) and integrated local area networks
(LANs) that appears as a single BSS to the logical
link control layer at any station associated with one
of those BSSs. The ESS can be thought of as the
coverage area provided by a collection of APs all
interconnected by the Distribution System. It may
consist of one or more IP subnets.
Station (STA): Any device that contains an IEEE 802.11 conformant
medium access control (MAC) and physical layer (PHY)
interface to the wireless medium (WM).
3. Deployment Architectures for Mobile IPv6 on 802.11
In this section, we describe the two most likely relationships
between Access Points (APs), Access Routers (ARs), and IP subnets
that are possible in an 802.11 network deployment. In this document,
our focus is mainly on the infrastructure mode [3] of 802.11.
Usually, a given STA is associated with one and only one AP at any
given instant; however, implementations are possible [4] where
multiple associations per STA may be maintained as long as the APs
are connected to disjoint DSs. An STA may be in communication with
an AP only when radio propagation conditions permit. Note that, as
with any layer-2 technology, handover from one layer-2 point of
attachment (AP) to another does not necessarily mean a change of AR
or subnet.
AR AR
AR | AR AR | AR
\ | / \ | /
Subnet 1 Subnet 2
/ / | \ \ / / | \ \
/ / | \ \ / / | \ \
/ | | | \ / | | | \
AP1 AP2 AP3 AP4 AP5 AP6 AP7 AP8 AP9 AP10
Figure 1. An 802.11 deployment with relay APs.
Figure 1 depicts a typical 802.11 deployment with two IP subnets,
each with three Access Routers and five Access Points. Note that the
APs in this figure are acting as link-layer relays, which means that
they transport Ethernet-layer frames between the wireless medium and
the subnet. Note that APs do not generally implement any particular
spanning tree algorithm, yet are more sophisticated than simple
bridges that would relay all traffic; only traffic addressed to STAs
known to be associated on a given AP would be forwarded. Each subnet
is on top of a single LAN or VLAN, and we assume in this example that
APs 6-10 cannot reach the VLAN on which Subnet 1 is implemented.
Note that a handover from AP1 to AP2 does not require a change of AR
(here we assume the STA will be placed on the same VLAN during such a
handoff) because all three ARs are link-layer reachable from an STA
connected to any AP1-5. Therefore, such handoffs would not require
IP-layer mobility management, although some IP-layer signaling may be
required to determine that connectivity to the existing AR is still
available. However, a handover from AP5 to AP6 would require a
change of AR, because AP6 cannot reach the VLAN on which Subnet 1 is
implemented and therefore the STA would be attaching to a different
subnet. An IP-layer handover mechanism would need to be invoked in
order to provide low-interruption handover between the two ARs.
Internet
/ | \
/ | \
/ | \
AR AR AR
AP1 AP2 AP3
Figure 2. An 802.11 deployment with integrated APs/ARs.
Figure 2 depicts an alternative 802.11 deployment where each AP is
integrated with exactly one AR on a disjoint VLAN. In this case,
every change of AP would result in a necessary change of AR, which
would require some IP-layer handover mechanism to provide for low-
interruption handover between the ARs. Also, the AR shares a MAC-
layer identifier with its attached AP.
In the next section, we examine the steps involved in any 802.11
handover. Subsequent sections discuss how these steps could be
integrated with an IP-layer handover mechanism in each of the above
deployment scenarios.
4. 802.11 Handovers in Detail
An 802.11 handover takes place when an STA changes its association
from one AP to another ("re-association"). This process consists of
the following steps:
0. The STA realizes that a handoff is necessary due to degrading
radio transmission environment for the current AP.
1. The STA performs a scan to see what APs are available. The
result of the scan is a list of APs together with physical layer
information, such as signal strength.
2. The STA chooses one of the APs and performs a join to
synchronize its physical and MAC-layer timing parameters with
the selected AP.
3. The STA requests authentication with the new AP. For an "Open
System", such authentication is a single round-trip message
exchange with null authentication.
4. The STA requests association or re-association with the new AP.
A re-association request contains the MAC-layer address of the
old AP, while a plain association request does not.
5. If operating in accordance with 802.11i [6], the STA and AP
would execute 802.1X EAP-on-LAN procedures to authenticate the
association (step 3 would have executed in "Open System" mode).
6. The new AP sends a Layer 2 Update frame on the local LAN segment
to update the learning tables of any connected Ethernet bridges.
Although we preface step 1 with step 0 for illustration purposes,
there is no standardized trigger for step 1. It may be performed as
a result of decaying radio conditions on the current AP or at other
times as determined by local implementation decisions. Some network
interface cards (NICs) may do scanning in the background,
interleaving scans between data packets. This decreases the time
required to roam if the performance of the current AP proves
unsatisfactory, but it imposes more of a burden on the AP, since
typically the STA places it in power-save mode prior to the scan,
then once the scan is complete, returns to the AP channel in order to
pick up queued packets. This can result in buffer exhaustion on the
AP and attendant packet loss.
During step 2, the STA performs rate adjustment where it chooses the
best available transmission rate. Rate adjustment can be quite
time-consuming as well as unpredictable.
Note that in some existing 802.11 implementations, steps 1-4 are
performed by firmware in rapid succession (note that even in these
implementations step 3 is sometimes performed in a host driver,
especially for newer implementations). This might make it impossible
for the host to take any actions (including sending or receiving IP
packets) before the handover is complete. In other 802.11
implementations, it is possible to invoke the scan (step 1) and join
(step 2) operations independently. This would make it possible to,
e.g., perform step 1 far in advance of the handover and perhaps in
advance of any real-time traffic. This could substantially reduce
the handover latency, as one study has concluded that the 802.11
beacon scanning function may take several hundred milliseconds to
complete [8], during which time sending and receiving IP packets is
not possible. However, scanning too far in advance may make the
information out-of-date by the time of handover, which would cause
the subsequent joint operation to fail if radio conditions have
changed so much in the interim that the target AP is no longer
reachable. So, a host may choose to do scanning based on, among
other considerations, the age of the previously scanned information.
In general, performing such subsequent scans is a policy issue that a
given implementation of FMIPv6 over 802.11 must consider carefully.
Even if steps 1 and 2 are performed in rapid succession, there is no
guarantee that an AP found during step 1 will be available during
step 2 because radio conditions can change dramatically from moment
to moment. The STA may then decide to associate with a completely
different AP. Often, this decision is implemented in firmware and
the attached host would have no control over which AP is chosen.
However, tools such as the host AP driver [10] offer full control
over when and to which AP the host needs to associate. Operation as
an Independent BSS (IBSS) or "ad-hoc mode" [3] may also permit the
necessary control, although in this latter case attachment to an
infrastructure AP would be impossible. Implementers can make use of
such tools to obtain the best combination of flexibility and
performance.
The coverage area of a single AP is known as a Basic Service Set
(BSS). An Extended Service Set (ESS) is formed from a collection of
APs that all broadcast the same ESSID. Note that an STA would send a
re-association (which includes both the old and new AP addresses)
only if the ESSID of the old and new APs are the same.
A change of BSS within an ESS may or may not require an IP-layer
handover, depending on whether the APs can send packets to the same
IP subnets. If an IP-layer handover is required, then FMIPv6 can
decrease the overall latency of the handover. The main goal of this
document is to describe the most reasonable scenarios for how the
events of an 802.11 handover may interleave with the message
exchanges in FMIPv6.
5. FMIPv6 Message Exchanges
An FMIPv6 handover nominally consists of the following messages:
a. The mobile node (MN) sends a Router Solicitation for Proxy
(RtSolPr) to find out about neighboring ARs.
b. The MN receives a Proxy Router Advertisement (PrRtAdv)
containing one or more [AP-ID, AR-Info] tuples.
c. The MN sends a Fast Binding Update (FBU) to the Previous Access
Router (PAR).
d. The PAR sends a Handover Initiate (HI) message to the New Access
Router (NAR).
e. The NAR sends a Handover Acknowledge (HAck) message to the PAR.
f. The PAR sends a Fast Binding Acknowledgement (FBack) message to
the MS on the new link. The FBack is also optionally sent on
the previous link if the FBU was sent from there.
g. The MN sends Fast Neighbor Advertisement (FNA) to the NAR after
attaching to it.
The MN may connect to the NAR prior to sending the FBU if the
handover is unanticipated. In this case, the FNA (step g) would
contain the FBU (listed as step c above) and then steps d, e, and f
would take place from there.
6. Beacon Scanning and NAR Discovery
The RtSolPr message is used to request information about the
router(s) connected to one or more APs. The APs are specified in the
New Access Point Link-Layer Address option in the RtSolPr and
associated IP-layer information is returned in the IP Address Option
of the PrRtAdv [2]. In the case of an 802.11 link, the link-layer
address is the BSSID of some AP.
Beacon scanning (step 1 from Section 4) produces a list of available
APs along with signal strength information for each. This list would
supply the necessary addresses for the New Access Point Link-Layer
Address option(s) in the RtSolPr messages. To obtain this list, the
host needs to invoke the MLME-SCAN.request primitive (see Section
10.3.2.1 of the 802.11 specification [3]). The BSSIDs returned by
this primitive are the link-layer addresses of the available APs.
Because beacon scanning takes on the order of a few hundred
milliseconds to complete, and because it is generally not possible to
send and receive IP packets during this time, the MN needs to
schedule these events with care so that they do not disrupt ongoing
real-time services. For example, the scan could be performed at the
time the MN attaches to the network prior to any real-time traffic.
However, if the interval between scanning and handover is too long,
the neighbor list may be out of date. For example, the signal
strengths of neighboring APs may have dramatically changed, and a
handover directed to the apparently best AP from the old list may
fail. If the handover is executed in firmware, the STA may even
choose a new target AP that is entirely missing from the old list
(after performing its own scan). Both cases would limit the ability
of the MN to choose the correct NAR for the FBU in step c during an
anticipated handover. Ongoing work in the IEEE 802.11k task group
may address extensions that allow interleaving beacon scanning with
data transmission/reception along with buffering at APs to minimize
packet loss.
Note that, aside from physical layer parameters such as signal
strength, it may be possible to obtain all necessary information
about neighboring APs by using the wildcard form of the RtSolPr
message. This would cause the current access router to return a list
of neighboring APs and would not interrupt ongoing communication with
the current AP. This request could be made at the time the MN first
attaches to the access router and periodically thereafter. This would
enable the MN to cache the necessary [AP-ID, AR-Info] tuples and
might enable it to react more quickly when a handover becomes
necessary due to a changing radio environment. However, because the
information does not include up-to-date signal strength, it would not
enable the MN to predict accurately the next AP prior to a handover.
Also, if the scale of the network is such that a given access router
is attached to many APs, then it is possible that there may not be
room to list all APs in the PrRtAdv.
The time taken to scan for beacons is significant because it involves
iteration through all 802.11 channels and listening on each one for
active beacons. A more targeted approach would allow the STA to
scan, e.g., only one or two channels of interest, which would provide
for much shorter interruption of real-time traffic. However, such
optimizations are currently outside the scope of 802.11
specifications.
7. Scenarios
In this section, we look at a few of the possible scenarios for using
FMIPv6 in an 802.11 context. Each scenario is labeled by the
sequence of events that take place, where the numbered events are
from Section 4 and the lettered events are from Section 5. For
example, "1abcde23456fg" represents step 1 from Section 4 followed by
steps a-e from Section 5 followed by steps 2-6 from Section 4
followed by steps f and g from Section 5. This is the sequence where
the MN performs a scan, then the MN executes the FMIPv6 messaging to
obtain NAR information and send a binding update, then the PAR
initiates HI/HAck exchange, then the 802.11 handover completes, and
finally the HAck is received at the PAR and the MN sends an FNA.
Each scenario is followed by a brief description and discussion of
the benefits and drawbacks.
7.1. Scenario 1abcdef23456g
This scenario is the predictive mode of operation from the FMIPv6
specification. In this scenario, the host executes the scan sometime
prior to the handover and is able to send the FBU prior to handover.
Only the FNA is sent after the handover. This mode of operation
requires that the scan and join operations (steps 1 and 2) can be
performed separately and under host control, so that steps a-f can be
inserted between 1 and 2. As mentioned previously, such control may
be possible in some implementations [10] but not in others.
Steps 1ab may be executed far in advance of the handover, which would
remove them from the critical path. This would minimize the service
interruption from beacon scanning and allow at least one
RtSolPr/PrRtAdv exchange to complete so that the host has link-layer
information about some NARs. Note that if steps ab were delayed
until handover is imminent, there would be no guarantee that the
RtSolPr/PrRtAdv exchange would complete especially in a radio
environment where the connection to the old AP is deteriorating
rapidly. However, if there were a long interval between the scan and
the handover, then the FBU (step c) would be created with out-of-date
information. There is no guarantee that the MN will actually attach
to the desired new AP after it has sent the FBU to the oAR, because
changing radio conditions may cause NAR to be suddenly unreachable.
If this were the case, then the handover would need to devolve into
one of the reactive cases given below.
7.2. Scenario ab123456cdefg
This is the reactive mode of operation from the FMIPv6 specification.
This scenario does not require host intervention between steps 1 and
2.
However, it does require that the MN obtain the link-layer address of
NAR prior to handover, so that it has a link-layer destination
address for outgoing packets (default router information). This
would then be used for sending the FNA (with encapsulated FBU) when
it reaches the new subnet.
7.3. Scenario 123456abcdefg
In this scenario, the MN does not obtain any information about the