HANDOVER OPTIMIZATION FOR HOST
AND NETWORK MOBILITY
Hai Lin, Houda Labiod
Ecole Nationale Supérieure des Tétécommunication (ENST),46 rue Barrault, 75634 Paris Cedex, France
Guo Zhi Wei, Anne Wei
University of Paris XII, 61 Avenue du General de Gaulle, 94010 Créteil Cedex, France
Keywords: Mobile IP, NEMO, handover, multihomed, multiple mobile routers.
Abstract: Mobile IP and NEMO (network mobility) provide continuous connectivity to the Internet to a node or a
mobile network when moving from one access router to another. Because of link switching delay and IP
protocol operation, packets destined to mobile nodes or mobile networks can be delayed or lost during the
handover period. This paper proposes solutions to improve the performance of handover in the context of
Mobile IPv6 and NEMO. We introduce a new control entity for both MIPv6 and NEMO to manage the
traffic between access routers and mobile nodes or mobile routers, to provide low-latency and low packet
loss for real-time services during the handover.
1 INTRODUCTION
The need of keeping connection with Internet in
everywhere and at every time is more and more
demanded in recently years. We can find two kinds
of mobility scenarios. In the first one, only mobile
node moves and attaches to different locations in the
Internet topology through immovable routers, using
MIPv4 (Perkins, 2002) or MIPv6 (Johnson, 2004)
mechanisms, we call it Host Mobility. The second
one concerns the scenario where a set of hosts move
collectively as a unit, we call it Network Mobility.
There are many situations where an entire network
might move and attach to the Internet topology from
“anywhere” at “anytime”, like a mobile network that
can be contained in a train or a ship. In both cases,
continuous connectivity must be supported. Hence,
recently, several extensions to MIPv6 have been
proposed aiming to reduce the handover latency and
packet loss to improve the handover performance for
real-time applications support.
In the case of network mobility, host mobility
support protocols may produce enormous signaling
which is not suited to the whole network movement.
Moreover, not all nodes in a large mobile network
may be sophisticated enough to run such mobility
support protocols. Network Mobility support
protocols have then been proposed in the context of
a recent IETF working group called NEMO. In
mobile networks, the weakest part comprises mainly
the mobile router. If the MR is down, all connections
between Internet and mobile node are disrupted.
Hence, multihomed architecture with multiple
mobile routers which offers multiple connections to
the Internet is proposed for mobile networks. This
architecture enables connections to be maintained
even if one of the mobile routers fails. Our work
related to handover optimization for network
mobility focuses on this specific NEMO
architecture.
The goal of this paper is to optimize handover
process both in host and network mobility but
considered separetely. Firstly, we propose a new
scheme to achieve MIPv6 fast handover by
introducing a component called Handover Control
Function (HCF) in Hierarchical Mobile IPv6. This
function allows us to predict a new attachment point
while a mobile node moves. Secondly, we enhance
the handover performance in the case of multiple
mobile routers installed in mobile network, by using
a new component called Intelligent Control Entity
(ICE).
The remainder of the paper is organized as
follows. Section 2 presents background and related
work of mobile IP and NEMO. Section 3 presents
our HCF Based Handover Function for a novel
MIPv6 scheme and the detailed protocol operation.
59
Lin H., Labiod H., Zhi Wei G. and Wei A. (2006).
HANDOVER OPTIMIZATION FOR HOST AND NETWORK MOBILITY.
In Proceedings of the International Conference on Wireless Information Networks and Systems, pages 59-64
Copyright
c
SciTePress
Section 4 deals with NEMO with multiple mobile
routers via a general architecture and a new scheme.
Finally, conclusion and future work are mentioned
in section 5.
2 BACKGROUND AND RELATED
WORK
As our paper deals with both host and network
mobility, we describe the several proposals related
to these two kinds of mobility separately.
2.1 Host Mobility with Mobile IPv6
Many kinds of wireless technologies, such as GSM,
UMTS, Wireless LAN, using the Host Mobility
architecture currently co-exist and are likely that their
number will further increase in the near future. R
ecently,
as the WLAN technologies, especially the IEEE
802.11 standards have got great attention, a growing
number of WLANs have been set up in public
buildings or corporate environments as access
networks to the Internet. In this paper, we focus on
improving the handover performance of Mobile
IPv6 over Wireless LAN.
Actually, the main proposals accepted by IETF
are Hierarchical Mobile IPv6 (HMIPv6) (Soliman
2005) and Fast Handover for MIPv6 (FHMIPv6)
(Koodli, 2005). HMIPv6 introduces Mobility
Anchor Point (MAP) (a special node located in the
network visited by a Mobile Node (MN)) who acts
somewhat like a local Home Agent (HA) for the
visiting MN. Moreover, HMIPv6 separates MN
mobility into micro-mobility (within one domain or
within the same MAP) and macro-mobility (between
domains or between MAPs). With this hierarchical
network structure, MAP can limit the amount of
signaling required outside the MAP's domain.
Therefore, the amount and latency of signaling
between a MN, HA and one or more Correspondent
Nodes (CNs) decrease.
FHMIPv6 reduces packets loss by providing fast
IP connectivity as soon as a new link is established.
The network uses layer 2 triggers to launch either
Pre-Registration or Post-Registration handover
scheme. In Pre-Registration scheme, the network
provides support for pre-configuration of link
information (such as the subnet prefix) in the new
subnet while MN is still attached to the old subnet.
By reducing the pre-configuration time on the new
subnet, it enables IP connectivity to be restored at
the new point of attachment sooner than would
otherwise be possible. In Post-Registration scheme,
by tunneling data between the previous point of
attachment and a new point of attachment, the
packets delivered to the old Care-of-Address (CoA)
are forwarded to the new CoA during link
configuration and Binding Update. So it is possible
to provide IP connectivity in advance contrarily to
the actual Mobile IP registration with the HA or CN.
Besides the main proposals, there have been
numerous approaches for providing lossless
handover and minimizing the handover delay. In
(Chaouchi, 2004), a Pre-Handover Signaling (PHS)
protocol is proposed to support the triggering of a
predictive handover and to allow the network to
achieve accurate handover decisions. In (Bi, 2004), a
Hierarchical Network-layer Mobility Management
(HNMM) framework is described in which an
integrated IP-layer handover solution is proposed to
provide optimized network connectivity. Also, a
Competition based Soft Handover Management
(CSHM) protocol (Kristiansson, 2004), and the
Multi-path Transmission Algorithm (Kashihara,
2002) are proposed to decrease packet loss during
handover.
2.2 Network Mobility
Network Mobility (NEMO) provides continuous
connectivity to the Internet to a set of nodes within a
mobile network. As illustrated in figure 1, a mobile
network is composed of one or more mobile IP-
subnets (NEMO-link) and is viewed as a single unit.
This network unit is connected to the Internet by
means of one or more Mobile Routers (MRs). Three
types of nodes behind the MR are defined : Local
Fixed Nodes, Local Mobile Nodes and Visiting
Mobile Nodes.
At a home link, an entity named Home Agent
(HA) is presented (figure 1), with which the mobile
router will register its care-of address and prefix.
While the mobile network is away from home, the
home agent intercepts packets on the home link
destined to the mobile network, encapsulates them,
and tunnels them to the MR’s registered care-of
address. At the foreign link, MRs get network layer
access to the global Internet from the Access
Router(s) (AR) (figure 1), via which packets from/to
Internet are transported.
The NEMO basic protocol (Devarapalli, 2005)
requires the MR to act on behalf of the nodes within
its mobile network. When an MR configures a new
CoA at a foreign link, it sends a Binding Update
message to its home agent, which contains its CoA
and its prefixes (in the case where the MR’s prefix
can be determined by home agent, prefix is not
included in Binding Update message). These
prefixes are then used by the HA to intercept packets
WINSYS 2006 - INTERNATIONAL CONFERENCE ON WIRELESS INFORMATION NETWORKS AND SYSTEMS
60
Figure 1: Architectural Components of NEMO.
addressed to the mobile nodes and these packets are
tunnelled to the MR. In response to the Binding
Update message, HA returns a Binding
Acknowledgement message to MR, which indicates
whether the HA has successfully processed the
Binding Update and has set up forwarding for the
Mobile network. If the procedure of Binding Update
is successful, a bidirectional tunnel between the MR
and the HA should be established. From now, the
MR encapsulates the packets from the mobile
network and tunnels them to HA, the latter
decapsulates the packets and forwards them to the
Correspondent Nodes (CNs). In the opposite
direction, the HA encapsulates the packets to the
mobile network and tunnels to the MR, the MR
decapsulates the packets and forwards them to the
mobile nodes.
As described above, the packets from/to mobile
network must go through the bidirectional tunnel,
which increase the packet’s size (add an additional
tunnel encapsulation) and the end-to-end delay. A
straight-forward approach to route optimization in
NEMO is proposed. Instead of preceding the
Binding Update with HA, MR sends Binding
Updates containing one or more Mobile Network
Prefix options to the CN. The CN having received
the Binding Update, can then set up a bi-directional
tunnel with the MR at the current care-of address of
the MR, and inject a route to its routing table so that
packets destined for addresses in the mobile network
prefix will be routed through the bidirectional tunnel
between the CN and the MR (Ng, 2005). But
establishing this kind of tunnel can be difficult to
perform, especially in the case where the CN
belongs to a mobile network node, and this mobile
network is away from its home link. However, we
can establish a tunnel to CN’s MR. To extend this
idea, if a tunnel can be established between the MR
and an entity which is located closer to the CN than
the HA, the route between mobile network and the
CN can be said to be optimized.
2.3 Open Research Issues
Besides the problems of handover delay and packet
loss in wireless LANs, there are many important
issues related to handover such as handover
architectures and handover decision algorithms that
should be studied. The handover architectures
concern mobility management and admission control
while the handover decision algorithms decide some
parameter adoptions and user preferences. In
addition of handover issues, quality-of-service
(QoS) and security are also the open research issues
that are addressed at current and future wireless
networks.
The most attractive feature for NEMO is the
simplicity of its NEMO basic protocol. Since it is an
extension of the Mobile IPv6 operation at the mobile
routers and home agents, the effective deployment
of mobile networks will strongly depend on how we
can overcome some critical issues such as sub-
optimal routing, handoff optimization, QoS
management, multihomed configurations, security
issues especially in nested configurations,
compatibility with Mobile IPv6, access control,
billing, and scalability. Some solutions have been
proposed to tackle some of these issues but some
other remain open and need further research like
handover, security and QoS support.
When moving from an AR to another, because of
a handover situation, traffic can be lost due to link
switching delay and IP protocol operation (CoA
configuration, Duplicate Address Detection, etc). Up
to now, there are no proposals and we aim to
propose some mechanisms to enhance NEMO
handover performance in the case of multiple MRs
installed in the mobile network (see section 4). This
type of mobile network is one kind of multihomed
mobile network; the latter can be configured into
serveral types (Ng, Paik, 2005): when a MR has
multiple egress interfaces, the mobile network has
multiple MRs, the mobile network is associated with
multiple HAs, and multiple global prefixes are
available in the mobile network. Advantages of such
architectures are listed in (Ernst, 2005) including
permanent and ubiquitous access, reliability, load
balancing, and preference settings.
3 HANDOVER CONTROL
FUNCTION (HCF) BASED
HANDOVER FOR MOBILE
IPV6
As mentioned in Section 2.1, MIPv6 handover has
been studied in some paper. We introduce a local
HANDOVER OPTIMIZATION FOR HOST AND NETWORK MOBILITY
61
intelligent entity called Handover Control Function
(HCF) which should be capable of controlling
several ARs and MNs.
3.1 HCF General Architecture
Linking with ARs, Handover Control Function
(HCF) enables to decide the MN’s new attachment.
We define four new messages: Handover Request
(HOReq), Handover Reply (HORep), Connection
Establish Information (CEInf) and Handover Finish
Confirmation (HFCon) messages in Mobile IPv6. It
is necessary to mention that HCF can manage
handover, resources distribution and security. Figure
2 illustrates the considered architecture.
Based on Figure 2, MN surveys periodically the
received signal strength. Once the signal strength
drops below the threshold predefined, MN will
begin to scan and discover the new available AP. It
reports to HCF the APs’ BSSID (Basic Service Set
Identifier) and signal strengths that it can probe.
Based upon the reported information, AR/AP’s
loading, and MN’s QoS demand, by using a
predefined algorithm, HCF decides whether or
which AP MN shall associate with and notifies MN
about the new AR/AP's information, such as AP's
BSSID, AR interface address, and sub-network
prefix. HCF decides which AR's interface MN
should move to as well. Consequently, the new
network prefix of MN will be notified by HCF
through HCFRep message accordingly.
The “IPv6 address allocation and assignment
policy" issued by RIPE (Ripe, 2003) provides the
guidelines for allocation and distribution of IPv6
addresses. This draft reveals that in an IPv6 access
network as MN moves across the subnets, the only
change in its IPv6 address occurs in subnet identifier
field of the address. The remaining portion of the
address, including 48 bit global routing prefix and
the 64 bit interface identifier remains unchanged.
Moreover, in our proposal, MN's interface identifier
is allocated according to the norm of EUI-64. It
ensures that the MN’s new CoA is unique in Mobile
Figure 2: HCF Based Handover for Mobile IPv6.
IPv6. Consequently, MN could configure its new
CoA and launches the Binding Update process even
if it is still attached with previous AR/AP. HCF also
knows MN's new CoA according to MN's old CoA
and MN's new network prefix. Furthermore,
Duplicated Address Detection (DAD) can be
omitted during handover.
After HCF sends the HCFRep message, HCF
will intercept all packets sent to the MN’s previous
CoA, and buffers these packets. MN will send CEInf
to HCF as soon as it finishes its handover process at
the layer 2. HCF then encapsulates these packets and
sends to MN with the HFCon message. Figure 3
shows messages exchange during the handover
procedure.
Figure 3: Protocol for HCF based handover.
3.2 HCF Procedure
HCF procedure is detailed as follows:
1) When MN registers to HA at the first time after
it attaches to a new AR/AP, it sends Binding Update
message to HA and HA responses with Binding
Acknowledgement.
2) Moving at sub-network, if the threshold of
received signal strength is overstepped, MN begins
to probe all neighbor AP's information, including
signal strength once. Then MN sends HCFReq
message directly to HCF to report the information of
its neighbor AP.
3) Receiving the HCFReq message, HCF decides
whether or which AR/AP MN shall associate with.
The choice of AR/AP is based mostly on the signal
strength that MN receives and AR/AP’s loading. For
example, if the number of registered MNs in one AR
or AP has reached a limit, HCF will not accept MN
to move to that network. After making the decision,
HCF sends the HCFRep message to MN. HCF
notifies MN about new AR/AP's information, such
as link prefix and AR's address. The information
WINSYS 2006 - INTERNATIONAL CONFERENCE ON WIRELESS INFORMATION NETWORKS AND SYSTEMS
62
will help MN to obtain a new CoA before it attaches
to the new AR/AP.
4) After MN receives the HCFRep message, it
knows that which AR/AP it will associate with and
will configure its new CoA. Once MN attaches to
the new AP and finishes the new CoA configuration,
it sends the CEInf message to HCF.
5) After receiving this message, HCF begins to
send the buffered packets to MN’s new CoA. Thus,
HCF sends the message HFCon to end the handover
procedure.
As shown in HCF procedure, a MN can obtain
its new CoA before it really attaches to the new
AR/AP, so handover is optimized to reduce
handover latency and packet loss.
4 HANDOVER FOR NEMO WITH
MULTIPLE MOBILE ROUTERS
To support handover optimization for network
mobility, we define a new architecture (figure 4). In
this architecture, we introduce a local Intelligent
Control Entity called ICE which should be capable
of controlling several ARs and the MRs attaching to
these ARs. In our case, the ICE can manage
handover, but it can also be used in the future for
managing connections, resources, QoS, security, etc.
An ICE domain contains an ICE and several ARs.
The ICE should possess the information of each its
ARs, like capacity, preferences, etc. Once an MR
attaches to an AR within an ICE domain, the ICE
should also collect this MR’s information, this
information is sent by the MR which knows the
ICE’s address by AR’s router advertisement
message. The information includes MR’s home
address, MR’s CoA, address of the attached AR,
capacity, preference, etc.
Our mechanism for handover optimization is
conceived for the case of multiple mobile routers
installed in mobile network. As illustrated in figure
4, if MRs are located separately in the mobile
network, as the mobile network moves, the MR at
the right end of the mobile network (MR1) will
perform handover firstly, when two others connect
always to previous AR (AR1). When the behind
MRs begin to perform handover, the MR at the right
end of the mobile network should complete its
handover. So in any time, it can exist at least one
MR which does not perform handover, and which
can transport the traffic addressed to the MRs being
performing handover. However, if the distance
between the MRs is not long enough, all MRs can
perform handover overlapping in time, so the
distance impacts the performance of handover
proposed by our mechanism.
When the MR1 (in figure 5) begins to perform
handover, as a response to some link-specific event
Figure 4: Our new NEMO architecture with 3 MRs.
(L2 “trigger”), which anticipates the handover, the
MR1 sends the handover initiate message to its ICE.
Upon receiving this message, the ICE knows that
MR1 will begin the handover, it should, according to
the collected information (capacity, preference, etc),
choose another MR which is at the same mobile link
and which is the best candidate to transport the
MR1’s traffic. If ICE can not find any candidate
MR, it returns a handover response message
indicating that the handover procedure has failed.
After having chosen the candidate MR (for example,
MR2 in figure 5), ICE sends a traffic transfer
message to MR1’s previous AR (AR1), to inform it
to transfer the packets addressed to MR1 to MR2.
From now on, AR1 encapsulates the packets
addressed to MR1 and sends them to MR2, the latter
decapsulates these packets and sends them to MR1
through the mobile link. To indicate whether the
procedure of establishment of the tunnel is achieved,
AR1 sends a handover response message to MR1. If
AR1 succeeds the establishment, this message must
contain the identity of the candidate MR (the
identity of MR2), this identity enables the MR1 to
know which MR transports its traffics. If not, the
MR1 knows that the procedure of establishment of
the tunnel has failed, and NEMO basic operation is
performed. If MR1 does not receive this message
over a certain time period, MR1 considers that the
establishment of the tunnel fails. In the case where
MR1 has disconnected to AR1, this message should
be sent to MR2, MR2 then transmits it to MR1.
Figure 5: MR1 performs handover.
HANDOVER OPTIMIZATION FOR HOST AND NETWORK MOBILITY
63
In the opposite direction, MR1 should send the
packets from the mobile network to Internet to MR2
until it completes the Binding Update, MR2 should
forward these packets to AR1. This procedure
enables all packets from the mobile network to
Internet to be transmitted without buffering even
MR1 at the handover state. Additionally, this
procedure ensures that packets are not dropped due
to ingress filtering. When MR1 has established the
connection with the new AR (AR2), i.e. it obtains
the new CoA, it involves the procedure of Binding
Update, which similarly as in NEMO. At the same
time, it performs the information registration
procedure as described above. Once the MR1
receives the Binding Update Acknowledgement
message, it stops transmitting the packets from
mobile network to Internet via MR2.
As described above, if it exists at least an MR at
non handover state (does not perform handover) at
any given time, the packets can always be
transported without being buffered during handover.
5 CONCLUSION AND
PROSPECTS
In this paper, we propose mechanisms to enhance
handover operation both in host and network
mobility. In the case of host mobility, HCF scheme
is proposed without modifying the part of classical
Access Router and Access Point in MIPv6. This new
HCF scheme allows MN to get the new CoA and to
launch Binding Update procedure before moving to
the new AR/AP. Moreover, the omission of DAD
process optimizes greatly handover performance.
Furthermore, by the means of buffering the traffic
during the layer 2 handover processes, then
resending to the new AR/AP attachment by HCF
after the layer 2 handover, the packet loss could be
minimized and overcome.
For NEMO mobility, our proposition provides
not only the minimization of delay and packet loss,
but takes into consideration the resource
management as well during handover. Additionally,
our protocol is compatible with the ingress filter
policy. However, since during handover, one MR
may transport other MRs’ traffics, the capacity of
the MRs should be designed to be large enough.
This paper details our primary concepts and
mechanisms. In the next step, our mechanisms will
be simulated and evaluated by using simulation
tools. In the future, we intend to investigate the other
MIPv6 or NEMO issues, such as cross-layer
solution, security aspects and QoS support taking
benefit of using the introduced new intelligent
entities to better manage the mobility and the
network resource.
ACKNOWLEDGEMENTS
This work is supported by the international project
PRA-SIP under Grant SIP 04-03.
REFERENCES
Perkins C., 2002. “IP Mobility support for IPv4,” RFC
3220.
Johnson D, Perkins C., Arkko J., 2004. "Mobility Support
in IPv6", RFC 3775.
Soliman H., 2005. "Hierarchical Mobile IPv6 mobility
management (HMIPv6)", RFC 4140.
Koodli R., 2005. "Fast Handovers for Mobile IPv6", RFC
4068.
Chaouchi H., 2004. “Pre-handover signaling for QoS
aware mobility management,” international journal of
network management 14, pp.367–374.
Bi Y., 2004. “An Integrated IP-layer Handover Solution
for Next Generation IP-based Wireless Network”,
Vehicular Technology Conference, Vol. 6 pp.3950 -
3954
Kristiansson J., 2004. "Application-layer Mobility support
for Streaming Real-time Media," Wireless
Communications and Networking Conference, Vol.1,
pp.268-273.
Kashihara S., 2002. "End-to-End Seamless Handover
using Multi-path Transmission Algorithm," In Internet
Conference, IC’02.
Devarapalli, V., 2005. "Network Mobility (NEMO) Basic
Support Protocol", RFC 3963.
Ng C., 2005. “Network Mobility Route Optimization
Solution Space Analysis”, draft-ietf-nemo-ro-space-
analysis-01 (work in progress).
Ng, Paik, 2005 “Analysis of Multihoming in Network
Mobility Support”, draft-ietf-nemo-multihoming-
issues-04 (work in progress).
Ernst T., 2005. "Goals and Benefits of Multihoming",
draft-multihoming-generic-goals-and-benefits-02
(work in progress).
Ripe, 2003. IPv6 Address Allocation and Assignment
Policy, ripe-267, January 2003.
http://www.ripe.net/ripe/docs/ipv6policy.html
WINSYS 2006 - INTERNATIONAL CONFERENCE ON WIRELESS INFORMATION NETWORKS AND SYSTEMS
64
