Sunteți pe pagina 1din 19

Wireless Pers Commun (2011) 59:217235

DOI 10.1007/s11277-009-9914-x
SMS-MIPv6: An End-to-End Mechanism for IPv6
Mobility Management in Mobile Networks
Hancheng Lu
Published online: 9 January 2010
Springer Science+Business Media, LLC. 2010
Abstract As IP has been extended from core networks to access networks, a mobile
network can be considered as an overlay of a traditional cellular network and an IP network.
SMS-MIPv6 attempts to integrate mobility management of these two kinds of networks.
The basic idea behind SMS-MIPv6 is to exploit existing mobility management in the cellular
network (i.e. in the formof well-dened short messages) to locate a Mobile Terminal (MT) in
the IPv6 network. We should emphasize that the motivation of SMS-MIPv6 is not to replace
or optimize existing mature mobility management schemes. On the contrary, as an entirely
end-to-end mechanism for IPv6 mobility management, it provides an alternative mechanism
for free peer-to-peer applications such as Voice over IP (VoIP) without support from mobile
network operators. We describe the implementation of SMS-MIPv6 in detail and analyze
its performance. The evaluation results show that SMS-MIPv6 achieves acceptable perfor-
mance so that it can be deployed in most current mobile networks. It performs best in terms of
signaling cost, data trafc overhead compared with Mobile IPv6 (MIPv6) and Proxy MIPv6
(PMIPv6). Moreover, SMS-MIPv6 can reduce the handover latency significantly, although
it is considered as a mobility management scheme for global mobility. However, it increases
the session initialization latency due to hybrid binding through the cellular network.
Keywords Mobility management Mobile networks IPv6 networks
1 Introduction
Nowadays, mobile networks provide subscribers with not only voice services, but also data
services. Requirements for such integrated services from subscribers facilitate IP to be
extended from core networks to access networks. It is also said that mobile networks are
being on their evolutional way to all-IP architectures. As a result, most Mobile Terminals
(MTs) are equipped with IP compatible wireless interfaces such as GPRS, EDGE beyond
H. Lu (B)
Department of Electronic Engineering and Information Science,
University of Science and Technology of China, 230027 Heifei, China
e-mail: hclu@ustc.edu.cn
123
218 H. Lu
GSM, CDMA. In some countries or areas, WiFi is also supported as a popular option for
Internet access. In this case, a mobile network is considered to consist of two overlay net-
works, i.e. a cellular network for traditional voice services and an IP network for emerging
data services.
Mobility management is necessary for services continuous as MTs roaming in the mobile
network. In the cellular network, a sophisticated and effective infrastructure [1] has already
been deployed and operated for mobility management for years, where two types of dat-
abases, i.e. the Home Location Register (HLR) and the Visitor Location Register (VLR), are
involved in tracking MTs. In the IP network, there are more uncertainties for the deployment
of mobility management. Many methods have been proposed. Among these, Mobile IPv6
(MIPv6) [2], a mobility enhancement to IPv6, has attracted most attention. As IPv6 is chosen
as the network layer for the next generation Internet, MIPv6 is also standardized by IETF
(Internet Engineering Task Force). Unfortunately, fewIP networks support MIPv6 until now.
The MIPv6 protocol not only requires modication of the IPv6 stack of the MT, but also
introduces a new entity, i.e. Home Agent (HA), to the network. Moreover, the recommended
route optimization mode [2] in MIPv6 could result in security problem due to the openness
of the IP network. Although a network-based localized mobility management scheme called
Proxy Mobile IPv6 (PMIPv6) [3] have been proposed to eliminate the participation of the
MT, most network operators still lack enthusiasm for deploying mobility management in
their IP networks. This will hinder MTs fromrunning IP based services such as Voice over IP
(VoIP), P2P (Peer-to-Peer), etc., which are becoming more and more popular in the mobile
environment. These services are end-to-end and require that MTs can track the locations of
each other for initializing IP communication and avoiding disruptions in IP communication
in spite of the movement.
Is it possible to integrate mobility management in the cellular network with that in the IP
network? SMS-MIP is an attempt to achieve this goal. The basic idea behind SMS-MIP is to
exploit existing mobility management in cellular networks to locate MTs in the IP network.
Actually, SMS-MIP is an end-to-end mechanism for mobility management in IP networks.
To implement SMS-MIP, only MTs are involved, and no network entity is introduced. There-
fore, it can be deployed easily by subscribers themselves as required. Moreover, security
of SMS-MIP can be enhanced by communicating condential information such as keys for
authentication through the close cellular network.
The rest of this paper is organized as follows. Section2 describes the motivation of
SMS-MIPv6. Section3 gives an overview of related work. Section4 describes the design
of SMS-MIP in detail. Two kinds of networks, i.e. the cellular network and the IP network,
are both involved. The performance of SMS-MIP is analyzed and evaluated in Sects. 5 and
6, respectively. Finally, Sect. 7 concludes this paper.
2 Motivation
We should emphasize that the motivation of SMS-MIPv6 is not to replace or optimize exist-
ing mature mobility management schemes (i.e. MIPv6, PMIPv6 and their optimized vari-
ations) in IPv6 networks. On the contrary, we provide an additional End-to-End mobility
mechanism for emerging IP-based Peer-to-Peer (P2P) applications (i.e. VoIP, streaming,
le sharing, etc.) when current network operators lack of incentive to deploy IP mobil-
ity management in their mobile networks. However, SMS-MIPv6 can be easily imple-
mented by re-using mobility signaling in MIPv6. With support from the network, it can
123
SMS-MIPv6: An End-to-End Mechanism 219
Mobile Network
WiFi
WiFi
WiMax
GPRS (IP)
GSM
Mobile Core
Network
WiFi
WiFi
WiMax
GPRS(IP)
GSM
MT1
MT2
Fig. 1 Application scenario of SMS-MIPv6
also be optimized with optimized variations of MIPv6 such as Fast handovers for MIPv6
(FMIPv6).
The most prominent characteristic of SMS-MIPv6 is that it can be controlled and imple-
mented by users themselves, just like a P2P software. As mobile networks are evolved to be
full IPv6 enabled, MIPv6 and PMIPv6 might be used.
An application scenario is shown in Fig. 1. Both MT1 and MT2 are equipped with cellular
interfaces (CIFs, i.e. GSM, CDMA, etc.) and IP compatible wireless interfaces (PIFs, i.e.
GPRS, WiFi, etc.). With existing mobility management deployed in the cellular networks,
MT1 and MT2 are always reachable through their CIFs by their phone numbers. However,
MT1 and MT2 want to initialize a low-cost VoIP session through their PIFs. Unlike the con-
stant phone numbers, MT1 and MT2 must vary their IP addresses as they move cross different
IP subnets. For example, In Fig. 1, MT2 is initialized in the GPRS network and handovers
to the WiFi network, which involves initialization and re-conguration of its IP address.
Therefore, tracking the IP address of a communication partner is crucial for initializing and
maintaining IP-based P2P sessions such as VoIP.
Actually, MT1 can obtain the IP address of MT2 at anytime, so does MT2. But the nec-
essary precondition is that an IP mobility management scheme has been deployed in the
network. Practically, this is impossible in most operated networks at present and near future.
Moreover, MTs might handovers among heterogeneous networks administrated by different
network operators, which introduces additional obstacles for the deployment of IP mobility
management.
SMS-MIPv6 is an attempt to track the IP address of an MT without support from net-
works. It can be entirely controlled and implemented on the MT, which facilitates IP-based
P2P applications in mobile networks. The basic idea behind SMS-MIPv6 is to utilize existing
mobility management architecture in the cellular network to locate an MT in the IP network.
3 Related Work
The work in this paper is related to mobility management schemes in the cellular network and
the IP network. However, to our best knowledge, there is no work that attempts to integrate
mobility management schemes of these two kinds of networks.
SMS-MIPv6 is an end-to-end mechanism for IPv6 Mobility Management as the mobile
network is evolving to the all-IP architecture. Besides SMS-MIPv6, many mobility manage-
ment schemes and corresponding optimizations are proposed by IETF for this goal, including
123
220 H. Lu
MIPv6 [2], Fast handovers for MIPv6 (FMIPv6) [4], Hierarchical MIPv6 (HMIPv6) [5] and
PMIPv6 [3].
MIPv6 [2] is a well-dened standard for global mobility in IPv6 networks. Mobility
operations in MIPv6 are described in Fig. 7b. In MIPv6, the HA performs mobility manage-
ment for MTs. Each MT is identied by a Home Address (HoA) in spite of its movement.
When an MT moves to a network, it performs movement detection by comparing the network
prex of the current network with that of its home network. If the MT nd itself in a foreign
network, it congures a Care-of-Address (CoA) based on the prex of the current network
and starts binding procedure by sending a Binding Update (BU) message to the HA. The BU
contains an association of the HoA with the CoA, which can be used to establish a binding
for the MT. After receiving the BU, the HA returns a Binding Acknowledgement (BACK)
message to the MT to indicate the binding status. If the binding is successful, a bi-directional
tunnel is established between the HA and MT for forwarding trafc from/to the HoA of the
MT. For convenience, any node communicating with a MT that is referred to in this paper
is dened as a Communication Partner (CP) of the MT. There are two possible modes for
communications between the MT and its CP, i.e. bidirectional tunneling mode and route
optimization mode [2]. In the bidirectional tunneling mode, the CP only knows the HoA of
the MT. Therefore, all trafc from/to the MT are forwarded through the bi-directional tunnel
between the MTand the HA. In the route optimization mode, the MTstarts binding procedure
to register its CoAwith the CP. In order to authorize the establishment of the binding, a return
routablility test must be also performed before binding procedure.
However, Mobility operations in MIPv6 often produce high latency and packet loss, which
are unacceptable for delay-sensitive and loss-sensitive applications. Therefore, additional
optimizations, i.e. FMIPv6 [4] and HMIPv6 [5], are proposed. The basic idea behind FMIPv6
is to accelerate handover procedures by exploiting link layer triggers [6]. With these link layer
triggers, the MT could start address conguration before the handover occurs and send pack-
ets as soon as it attaches to a new link. Therefore, the handover latency can be reduced
significantly. HMIPv6 is a localized mobility enhancement to MIPv6. In HMIPv6, a network
entity called Mobility Anchor Point (MAP) is introduced, which is responsible for mobility
in the local domain. When the MT moves between different subnets in the same network
domain, it performs binding procedure with the MAP instead of the HA. Thus, when the MT
moves to a network domain far away from its home network, both the handover latency and
signaling messages across backbone networks are reduced significantly.
Like HMIPv6, PMIPv6 [3] is proposed for local mobility in a network domain. However,
its also a network-based mobility management scheme. That means mobility management
is transparent to the MT. As no support from the MT is required, PMIPv6 is more convenient
for deployment. Mobility operations for PMIPv6 are described in Fig. 7c. In PMIPv6, two
network entities, i.e. Mobile Access Gateway (MAG) and Local Mobility Anchor (LMA)
are introduced for local mobility. When the MAG detects an event that an MT is attached to
its access link, it sends a Proxy Binding Update (PBU) message to the LMA. If the PBU is
acceptable, the LMA sends a Proxy Binding Acknowledgement (PBACK) message includ-
ing a Home Network Prex (HNP) for the MT. Upon receiving the PBACK, the MAG sends
a Router Advertisement (RA) message to the MT. At the same the time, a bi-directional
tunnel is established between the MAG and the LMA, which is used for forwarding trafc
from/to the MT. The MT extracts the HNP from the RA and congures its address using
either stateful or stateless address conguration modes. The HNP is specied to the MT
and it keeps unchanged as the MT moves between different networks in the same domain.
Therefore, from the perspective of the MT, the entire network domain appears as a single
link. The network ensures that the MT is not aware of mobility with respect to its IP layer.
123
SMS-MIPv6: An End-to-End Mechanism 221
4 Proposed Scheme
Fromthe viewpoint of SMS-MIPv6, a mobile network consists of two types of networks, i.e.
the cellular network and the IP network. Each MT is able to connect to these two networks
simultaneously. For simplicity, an MT is assumed to be equipped with two interfaces, one for
the cellular network (CIF) and the other for the IP network (PIF). For many MTs in China,
the CIF is GSM, and the PIF is GPRS. However, other types of interfaces such as CDMA,
EDGE, WiFi can also be supported in SMS-MIPv6.
End-to-end voice is a major service in the cellular network. Compared with traditional
voice through the CIF, VoIP through the PIF has the advantage of lower cost. And it tends
to be a popular service as IP is extended to MTs at the edge of the mobile network. SMS-
MIPv6 is designed to facilitate the application of such kind of end-to-end service in the
mobile environment. It is easy to be implemented without deploying any network entity. In
SMS-MIPv6, address mapping and hybrid binding are proposed to support mobility as that
in MIPv6, which involve not only the IP network, but also the cellular network.
(1) In MIPv6, each MT is assigned a unique HoA for IP-based communication in spite of
its movement. In SMS-MIPv6, there is no network entity for allocating and maintaining
the HoA for the MT. Instead, each MT forms its unique HoA by address mapping from
its phone number.
(2) In MIPv6, the HA is used to track current locations of MTs in the IP network. That
means an MT is always reachable through the HA. In SMS-MIPv6, there is no network
entity for this function. Instead, each MT delivers information about its current location
in the IP network to its communication partner through the cellular network as well as
the IP network, which is called hybrid binding in this paper.
4.1 Address Mapping
From the view point of a subscriber, an MT is usually identied by its phone number in the
mobile network. The phone number is used as a unique address for the MT in the cellular
network. And it keeps unchanged as the MT moves. In the IP network, the MT is addressed
by its IP address, whose prex identies the currently located subnet based on routing.
Therefore, as the MT moves between different subnets, its IP address should be changed
accordingly.
To preserve IP communication, each MT in SMS-MIPv6 is assigned two IPv6 addresses,
an HoA for identication and a CoA for packet routing, which is similar to that in MIPv6.
The CoA is allocated by the subnet that MT is currently attached to. However, the HoA
is formed by address mapping as shown in Fig. 2. It is a link-local address for identica-
tion in the IPv6 network, and its unique is guaranteed by the unique of the MTs phone
number.
The addressing model in SMS-MIP is shown in Fig. 2. Generally, if possible, a subscriber
prefers VoIP through the PIF to traditional voice through the CIF for achieving lower cost.
But he/she only knows the phone number of his/her CP. In SMS-MIPV6, First, the subscriber
inputs the corresponding phone number in the MT. By address mapping, the phone num-
ber is converted to the HoA. For transport protocols and services above the IP layer, they
only need to know the HoA. But for IPv6 packets send from/to MT, the CoA is used as
the source/destination address, and the HoA is included in the Home Address destination
option or type 2 routing header added to these IPv6 packets. In this way, the use of the CoA
123
222 H. Lu
Subscriber1 Subscriber2
Services
Transport
IP/MIPv6
Link
Services
Transport
IP/MIPv6
Link
CoA
Src=CoA
Dest. Option: HoA
Dst=CoA
Type 2 Routing Header: HoA
HoA
Phone Number
Address Mapping
HoA = FE80:: + Phone Number
MT2
Callee
MT1
Caller
HoA: Home Address
CoA: Care-of Address
Fig. 2 Addressing model in SMS-MIPv6
Fig. 3 State transition for hybrid
binding
SMS
Binding
IP
Binding
Send/Receive
SMS-BU
Key Updating
S
e
n
d
/
R
e
c
e
i
v
e
I
P
-
B
U
is transparent to transport protocols and services above the IP layer. So does the movement
in the IP networks.
4.2 Hybrid Binding
Hybrid binding consists of SMS binding through the cellular network and IP binding through
the IP network. It is used to bind the phone number with the IP address for the MT, where
two kinds of Binding Update (BU) messages, i.e. SMS-BUand IP-BU, are dened. SMS-BU
messages are delivered to MTs by exploiting mobility management in the cellular network.
For each MT, the state transition for hybrid binding is shown as Fig. 3.
Initially, an MT is in the SMS binding state. To start IP-based communication such as
VoIP, the MT sends the SMS-BU via Short Message Service (SMS) [1] to its CP with infor-
mation on its IP conguration. In SMS-MIPv6, binding is bidirectional. If the sSMS-BU is
accepted, MT will also receive a SMS-BU from its CP.
SMSis the most widely used data service in the current cellular network for the interchange
of short messages between MTs. It can be considered as an end-to-end service, although there
are SMS Centres (SMSCs) forwarding and storing short messages for MTs. Moreover, SMS
depends on the mobility management infrastructure consisting of HLRs and VLRs in the
cellular network to locate an MT and then deliveries short messages to the MT.
The SMS-BU is a short message that identies its destination by the phone number and
contains an association of the HoAwith the CoAfor the MT. By setting the User Data Header
(UDH), a SMS-BU can be distinguished from a simple text message, which is processed by
the SMS-MIPV6 module instead of displayed as text on the MT. The format of the SMS-BU
is shown in Fig. 4. The Acknowledge (A) bit is set to indicate that the SMS-BUis also used as
an acknowledgement message. In SMS-MIPv6, binding is bidirectional. In the SMS binding
state, SMS-BUmessages are exchanged between MTs. When Abit is set, the status eld will
exist and it indicates the process result of the SMS-BU previously received by the MT. Also
123
SMS-MIPv6: An End-to-End Mechanism 223
Fig. 4 Format of SMS-BU
message
there is a key option for authentication for the following IP binding when K bit is set. The
close cellular network is more secure for transmitting such kind of condential information
than the open IP network. Moreover, certain security options have already been proposed in
the technical specication for SMS [7].
However, SMS is a non-real-time system. That means SMS-BU messages might suf-
fer high delay when the SMS system is busy. In Sect. 5, the analysis shows that this delay
would affect the session initialization latency significantly. However, the session initiali-
zation latency only occurs when a call or connection is setting up. It does not contribute
to the handover latency during the call or connection. To reduce the session initialization
latency, extensions are required for the SMS system. For example, if the differentiated ser-
vice is supported, each SMS-BU message can be set to a higher priority than general SMS
messages.
After sending or receiving the SMS-BU, the MT enters the IP binding state. In this state,
the binding procedure is the same as that described in MIPv6. The MT exchanges the IP-BU
with its CP periodically to create or maintain the corresponding binding cache entry. The IP-
BU is sent through the IP network and authenticated with the key exacted from the SMS-BU,
which format is the same as that of the BU in MIPv6.
For security, each key used for authenticating the IP-BU is associated with a lifetime
dened in the key option as shown in Fig. 4. When the lifetime expires, the MT enters the
SMS binding state and begins to send the SMS-BU with the key option to update the key.
4.3 Implementation
SMS-MIPv6 is easy to be implemented on MTs with software update. No special hardware
support is required. We have already implemented a prototype of SMS-MIPv6 in Google
Android. In SMS-MIPv6, the MT that initializes the binding procedure is dened as a binder,
and its CP is dened as a bindee. In SMS-MIPv6, binding is bidirectional. Therefore, both
binder and bindee should be implemented on the MT for end-to-end mobility management.
Figure 5 shows the implementation of the binder and the bindee. As a binder, the MT
forms its HoA by address mapping with its phone number. It also informs bindee of its CoA
by hybrid binding, where the SMS-BU is used to deliver the association of the HoA with the
CoA before IP-based communication is established and negotiate the key for authentication
for the IP-BU. As a bindee, the MT creates and maintains a binding cache entry for binder as
it does in MIPv6, except that it might receive BU messages from both the cellular network
(i.e. SMS-BU) and the IP network (i.e. IP-BU). In SMS-MIPv6, the binding procedure is
bidirectional. When receiving the SMS-BU, the MT always sends the SMS-BU with A bit
set as an acknowledgement. The rst SMS-BU with A bit set will create an entry in binding
cache on the binder for the bindee. Then the binder could start the IP binding procedure with
the bindee.
123
224 H. Lu
Active PIF
Get IP
address?
Form HoA by
Address Mapping
Send SMS-BU
Receive
SMS-BU?
(A=1)
N
Y
N
Retry Limit
Exceeded?
Support VoIP?
Send IP-BU
Binding List
Expires?
Retry Limit
Exceeded?
Receive
IP-BACK?
Key Expires?
Send SMS-BU
Receive
SMS-BU?
(A=1)
Retry Limit
Exceeded?
Create an Entry in
Binding List
Traditional Voice over CIF
N
Y
Y
Update
Binding List
N
Y
Update Binding
List and Key
Y
N
Y
N
N
Y
N
Y
N Y
Y
N
PIF
Actived?
Y
N
Receive SMS-BU
Active PIF
Get IP
address?
Send SMS-BU
(A=1)
N
Y
N Find an Entry in
Binding Cache?
Create an Entry in
Binding Cache
Receive IP-BU
PIF Actived?
Update
Binding Cache
Y
Send IP-BACK
N Traditional
Voice over CIF
Y
Form HoA by
Address Mapping
(a)
(b)
Fig. 5 Implementation of SMS-MIPv6
5 Performance Analysis
Besides the proposed SMS-MIPv6, two popular mobility management schemes in IPv6 net-
works, i.e. MIPv6 and PMIPv6, are also analyzed in this section. Either SMS-MIPv6 or
PMIPv6 can be considered as an enhancement to MIPv6, as they can be easily implemented
by re-using signaling messages fromMIPv6. PMIPv6 can be considered as a localized mobil-
ity optimization for MIPv6, while SMS-MIPv6 can be considered as an end-to-end extension
for MIPv6. Optimized variations of MIPv6 (i.e. FMIPv6, HMIPv6) are not analyzed for the
following reasons:
(1) SMS-MIPv6 is not proposed to optimize or replace MIPv6, it just provide an alternative
end-to-end mobility management scheme for subscribers.
(2) With support fromthe network, FMIPv6 can also be easily integrated with SMS-MIPv6.
However, its performance depends on link layer technologies.
(3) Both HMIPv6 and PMIPv6 can be considered as localized mobility optimization for
MIPv6. In the analysis, they achieve similar performance.
Generally, two kinds of metrics are used to assess the performance of mobility manage-
ment schemes. One is used to indicate cost for network operators, such as mobility signaling,
data overhead, the other is used to quantify Quality of Service (QoS) for subscribers, such as
session initialization latency, handover latency. In this paper, these metrics are all analyzed
for MIPv6, PMIPv6 and SMS-MIPv6. Specifically, in the following analysis, we assume that
there is no latency for packet/message processing on network entities or MTs.
5.1 System Model
5.1.1 Mobility Model
This paper assumes a mobile network architecture with two tiers of access points for MTs.
Base Stations (BSs) cover hexagonal cells and form the rst tier for cellular access. Access
123
SMS-MIPv6: An End-to-End Mechanism 225
Session
Start
AR1
AR2
AR3
ARi
t'
r1
t
r2
t
r3
t
ri
t
1
t
s
Travel
Path
Fig. 6 Travel path and AR residence times
Routers (ARs) connects several BSs and form the second tier for IP access. It also can be
considered that each ARcovers several cells, as shown in Fig. 6. Amechanismfor IP mobility
management is required when and only when MTs move across different ARs.
Modeling the probability distribution of the number of handovers during an IP-based
session plays a significant role in performance analysis for the mobile network. Consider a
scenario that an MT travels a path that consists of ARs. As the MT traverses ARs, its resi-
dence times (t
r1
, t
r2
, t
r3
, . . .) are assumed to be independent, identically distributed random
variables. At time t
1
, the MT initializes the IP-based session so that the remaining residence
time in the AR1 is t

r1
. Let N
h
be the number of handovers per session. Given the MTs AR
residence times, N
h
depends on the session holding time t
s
. Hence,
Pr
_
N
h
= n

r1
, t
r2
, t
r3
, . . .
_
=

Pr
_
t

r1
< t
s
< t

r1
+t
r2
_
n = 1
Pr
_
t

r1
+
n

i =2
t
ri
< t
s
< t

r1
+
n+1

i =2
t
ri
_
n > 1
In the analysis, we assume that the session arrival process follow the Poisson distribution
in a given time interval with rate . In this case, it can be deduced that t
s
is exponentially
distributed with mean 1/. We are interested in E (N
h
), which is given by
E
_
N
h

r1
, t
r2
, t
r3
, . . .
_
=

n=1
n Pr
_
N
h
= n

r1
, t
r2
, t
r3
, . . .
_
= e
t

r1
_
1 +

i=2
e

i
k=2
t
rk
_
(1)
For the MT, the ARresidence time t
ri
(i = 1, 2, 3, . . .) is assumed to followan exponential
distribution with mean 1/. From (1), we can obtain that
E(N
h
) =

(2)
In [8], it also shows that (2) holds for arbitrary session holding time distributions and
stationary AR residence time distributions.
5.1.2 Network Model
As shown in Fig. 7a, the network model used for analysis consists of three network domains.
In the visit domain, an MT can access the network through the CIF or PIF. For PMIPv6,
the visit domain is also dened as LMA domain. Moreover, the AR is used as MAG and
123
226 H. Lu
Visit/LMA
Domain
CP
BS
AR/
MAG
BR/
LMA
AAA
HA
MT: Mobile Termination BS: Base Station
CP: Communication Partner AR: Access Router
MAG: Mobile Access Gateway HA: Home Agent
LMA: Local Mobility Anchor BR: Border Router
MT
Home
Domain
CP
Domain
Network Model
Link
Down
Link
Up
Address
Acquisition
BU
BACK
HoTI
CoTI
HoT
CoT
Movement
Detection
Address Configuration
HA Binding
CN Binding
Binding Refresh
t
d
t
a
t
br
t
ha
T
L2
T
L3
Link layer
handover Latency
IP layer handover
Latency
BU
BACK
(with Return Routability)
t
cn
HoTI
CoTI
HoT
CoT
BU
BACK
MIPv6 Signaling sequence on MT
Detached
Event
Attatched
Event
PBU
PBACK
Rtr Adv
Local Binding Binding Refresh
t
p
t
br
T
L2 T
L3
Link layer
handover Latency
IP layer handover
Latency
PBACK
PBU
PMIPv6 Signaling sequence on AR/MAG
Link
Down
Link
Up
Address
Acquisition
SMS-BU
SMS-BU
(A=1)
IP-BU
IP-BACK
SMS-BU
SMS-BU
(A=1)
Movement Detection
Address Configuration
SMS Binding
IP Binding
Key Updating
t
d
t
a
t
cell
t
ku
t
br
T
L2
T
L3
Link layer
handover Latency
IP layer handover
Latency
SMS-MIPv6 Signaling sequence on MT
(a)
(b)
(d)
(c)
Fig. 7 Network model and mobility signaling sequences
the Border Router (BR) is used as LMA. All trafc from/to the outside of the LMA domain
should pass through the LMA.
Based on the network model, signaling sequences for different mobility management
schemes are analyzed in Fig. 7b, c, d. Both SMS-MIPv6 and MIPv6 support global mobility
among different network domains. They are also considered as host-based mobility man-
agement schemes, where the MT must be involved. Therefore, signaling sequences for
SMS-MIPv6 and MIPv6 are described from the view point of the MT. On the contrary,
PMIPv6 can only be used for local mobility management within a network domain. It is con-
sidered as a network-based mobility management scheme, where the MAGperforms mobility
signaling instead of the MT. Therefore, the signaling sequence for PMIPv6 is described from
the view point of the MAG.
5.2 Signaling Cost
In this paper, the signaling cost is dened as the total number of signaling messages used
for mobility management during the session holding time. These messages are exchanged
between MTs or an MT and a network entity (e.g. HA), which are denoted as C
M
and C
N
,
respectively.
In MIPv6, an MT registers its current IP address with the HA in spite of its session status.
Signaling messages such as BUs are sent for updating location information when a handover
is involved or for refreshing location information (i.e. binding refreshing) at a given interval
t
MIPv6
br
. The average rate for refreshing location information while the MT still stays in the
coverage of an AR is
_
1/t
MIPv6
br
_
, where x is the integer part of a real number x. For
123
SMS-MIPv6: An End-to-End Mechanism 227
simplicity, let L
S
denote the average length of the signaling message. Therefore, C
MIPv6
N
can
be expressed as
C
MIPv6
N
=
E (N
h
)
_
C
MIPv6
ho
+
_
1
t
MIPv6
br
_
C
MIPv6
r f
_
L
S
,
where C
MIPv6
ho
and C
MIPv6
r f
denote handover signaling cost and refreshing signaling cost,
respectively. Let L
M
denote the length of signaling message M. According to Fig. 7b, we
have
C
MIPv6
ho
= C
MIPv6
r f
= L
BU
+ L
BACK
Therefore, C
MIPv6
N
is
C
MIPv6
N
=
E (N
h
)
_
1 +
_
1
t
MIPv6
br
__
(L
BU
+ L
BACK
)
L
S
= 2E (N
h
)
_
1 +
_
1
t
MIPv6
br
__
(3)
C
M
can be obtained in a similar manner, except that the MT must perform a return routa-
bility procedure with its CP to create a binding management key for authorizing the BU and
BACK. Therefore, C
M
can be expressed as
C
MIPv6
M
=
E (N
h
)
_
1 +
_
1
t
MIPv6
br
__
(L
rr
+ L
BU
+ L
BACK
)
L
S
= 6E (N
h
)
_
1 +
_
1
t
MIPv6
br
__
,
where L
rr
= L
HoT
+ L
HoT1
+ L
CoT
+ L
CoT1
= 4L
S
. The total signaling cost for MIPv6
can be obtained as
C
MIPv6
s
= C
MIPv6
N
+C
MIPv6
M
= 8E (N
h
)
_
1 +
_
1
t
MIPv6
br
__
(4)
In PMIPv6, the movement of the MTis transparent to its CP. Signaling messages for mobil-
ity management are only exchanged between the MT and LMA. That means C
PMIPv6
M
= 0.
According to Fig. 7c, the total signaling cost for PMIPv6 can be expressed as
C
PMIPv6
s
= C
PMIPv6
N
=
E (N
h
)
_
1 +
_
1
t
PMIPv6
br
__
(L
PBU
+ L
PBACK
)
L
S
= 2E (N
h
)
_
1 +
_
1
t
PMIPv6
br
__
, (5)
where t
PMIPv6
br
denotes the interval for binding refreshing in the local LMA domain. We can
see that PMIPv6 can reduce the signaling cost significantly compared with MIPv6. How-
ever, PMIPv6 is used for localized mobility management as MTmoves within the same LMA
domain. When the MTmoves between different LAMdomains, global mobility management
protocols such as MIPv6 should be involved. In this case, the total signaling cost is increased
significantly, which can be expressed as
123
228 H. Lu
C
PMIPv6
s
= 2E (N
h
)
_
1 +
_
1
t
PMIPv6
br
__
+
8E (N
h
)
N
R
_
1 +
_
N
R
t
MIPv6
br
__
, (6)
where N
R
denotes the number of ARs in an LMA domain.
SMS-MIPv6 implements end-to-end mobility management, where no network entities are
involved. That means C
SMS-MIPv6
N
= 0. However, SMS-MIPv6 involves additional signaling
cost in the cellular network compared with MIPv6 and PMIPv6. The total signaling cost can
be expressed as
C
SMS-MIPv6
s
= C
cell
s
+C
IP
s
(7)
Let t
ku
= t
SMS-MIPv6
br
, where t
SMS-MIPv6
br
denotes the interval for IP binding refreshing.
According to Fig. 7d, we have
C
cell
s
=
_
1
t
ku
_
(L
SMS-BU
+ L
SMS-BACK
)
L
S
= 2
_
1
t
SMS-MIPv6
br
_
(8)
C
IP
s
=
E (N
h
)
_
1 +
_
1
t
SMS-MIPv6
br
__
(L
IP-BU
+ L
IP-BACK
)
L
S
= 2E (N
h
)
_
1 +
_
1
t
SMS-MIPv6
br
__
(9)
5.3 Data Trafc Overhead
During the session holding time, the data trafc overhead is composed of tunnel cost C
t un
and header cost C
hdr
In mobility management schemes, IPv6-in-IPv6 tunnels are usually
established for redirecting data packets to/from the current location of MTs. In this case,
C
t un
is involved as an additional IPv6 header is added to each tunneled data packets. On the
other hand, IPv6 extension headers (i.e. destination option header, routing header) are added
to each packet sent from/to MTs, which provide transport protocols and services above the
IP layer services with the HoA information so that the change of the CoA is transparent. This
kind of cost is denoted as C
hdr
.
In MIPv6, route optimization is used as default mode for data packet transmission between
an MT and its CP. However, before the MT registers its CoA with its CP for the rst time,
data packets should be delivered through the tunnel established between the MT and the HA.
Let be the data packet arrival rate in unit of packet per time and L
H
ext
be the length of header
ext. Then, we have
C
MIPv6
t un
= t
cn
L
H
IPv6
,
where t
cn
is the average latency for the MT to perform the binding procedure with its CP.
After successful binding, the MT sends data packets to its CP directly. In this case, an IPv6
extension header (i.e. destination option header) containing the HoA information is added to
each data packet sent from the MT. Therefore, C
MIPv6
hdr
is given by
C
MIPv6
hdr
= L
H
dst opt
The total data trafc overhead in MIPv6 is
C
MIPv6
d
= C
MIPv6
t un
+C
MIPv6
hdr
=
_
t
cn
L
H
IPv6
+ L
H
dst opt
_
(10)
123
SMS-MIPv6: An End-to-End Mechanism 229
In PMIPv6, a tunnel is established between the MAG and LMA. Data packets from/to the
MT are always delivered through this tunnel. C
PMIPv6
t un
can be expressed as
C
PMIPv6
t un
= L
H
IPv6
As the movement of the MT is transparent to its CP, C
PMIPv6
hdr
is always equal to 0. There-
fore, the total data trafc overhead in PMIPv6 is
C
PMIPv6
d
= C
PMIPv6
t un
= L
H
IPv6
(11)
However, PMIPv6 can only support localized mobility within the same LMA domain. If
the MT moves between different LMA domains, MIPv6 should be also supported. In this
case, the total data trafc overhead is increased significantly, which can be obtained as
C
PMIPv6
d
= C
MIPv6
d
+C
PMIPv6
d
=
_
t
cn
L
H
IPv6
+ L
H
dst opt
+ L
H
IPv6
_
(12)
In SMS-MIPv6, data packets are transmitted between the MT and its CP directly after
hybrid binding. There is no tunnel cost. Therefore, the total data trafc overhead is
C
SMS-MIPv6
d
= C
SMS-MIPv6
hdr
= L
H
dst opt
(13)
5.4 Session Initialization Latency
In the mobile network, an MT varies its IP address as it moves across different ARs. There-
fore, it needs some time to obtain the current IP address of a given MT through mobility
management before any IP communication can be established. The session initialization
latency (T
i ni t
) is used to indicate such kind of latency. It is dened as the delivery latency of
the rst IP data packet that is sent from an MT to its CP when an IP session is initialized.
In the wired network, the transmission latency between node A and node B is assumed to
be proportional to the hop number between node Aand node B. Therefore, it can be expressed
as t
AB
= d
AB
, where is the average unit transmission latency over the wired link.
Specially, the average transmission latency between an MT and its attached AR is dened as
( > 1), where a wireless link is also involved.
In MIPv6, each MT sends BU messages to the HA to registers its current IP address as it
is attached to a new AR. Therefore, the MT can always send the rst data packet to its CP
through the HA. We have
T
MIPv6
i ni t
= ( +d
ARBR
+d
BRHA
+d
HACP
) (14)
In PMIPv6, the MAG (i.e. AR) instead of the MT sends BU messages to the LMA for
registration. Therefore, all packets to or from the MT must be delivered through the LMA.
We have
T
PMIPv6
i ni t
= ( +d
MAGLMA
+d
LMACP
) (15)
Unlike MIPv6 and PMIPv6, SMS-MIPv6 involves no network entity during mobility
management. Moreover, SMS-MIPv6 is an on-demand protocol coexisting with IP sessions.
That means the MT performs hybrid binding with its CP when and only when there is an IP
session between these two entities. Therefore, SMS-MIPv6 can reduce signaling overhead.
On the other hand, it increases the session initialization latency, which can be expressed as
T
SMS-MIPv6
i ni t
= t
cell
+ ( +d
ARBR
+d
BRCP
) , (16)
123
230 H. Lu
Fig. 8 Round trip time for SMS
in cellular network
where t
cell
is a round trip time between the MT and its CP as SMS binding is performed
in the cellular network. Fig. 8 shows t
cell
in the mobile network operated by China Mobile
Communication Corporation. The experiment data is obtained on 21 February, 2009.
1
We
can see that t
cell
is below 10s in most cases. However, SMS is a non-real-time system, t
cell
might increase significantly when the system is busy on some holidays. To achieve smaller
t
cell
, messages for SMS binding should be transmitted and processed with higher priority
than general short text messages. In this case, the SMS system should be extended to support
the differentiated service.
5.5 Handover Latency
When an MT moves between two ARs, it performs link layer (L2) handover as well as IP
layer (L3) handover to sustain existing IP connectivity. In this paper, the latency involved in
L3 handover (T
L3
) is focused on.
In MIPv6, according to Fig. 7b, T
MIPv6
L3
consists of delay due to movement detection
(t
d
), IP address conguration (t
a
) and binding update procedure (t
ha
+t
cn
), which can be
expressed as
T
MIPv6
L3
= t
d
+t
a
+t
ha
+t
cn
(17)
To detection movement, the MT can send a Router Solicitation (RS) message to request a
RA message from its attached AR. The RA provides MT with the prex information of the
subnet. Therefore, we have
t
d
= 2
From the prex information, the MT forms its CoA by using IPv6 stateless or stateful
address conguration. To validate the uniqueness of the CoA, the MT performs duplicate
address detection (DAD) process. According to [9], t
a
is
t
a
= R D,
where R and D denote RetransTimer and DupAddrDetectionTransmits specied in [10],
respectively.
1
In the experiment, eight students collected 240 short messages. There were 120 remote messages, which
were delivered between users in different cities, and 120 local messages, which were delivered between users
in the same city. The results showed that t
cell
for local messages was a little larger than that for remote
messages. However, the non-real-time characteristic of SMS would have more contribution to the variation of
t
cell
.
123
SMS-MIPv6: An End-to-End Mechanism 231
As the latency due to message process is ignored, t
ha
and t
cn
are determined by the
transmission latency. Therefore, we have
t
ha
= 2 ( +d
ARBR
+d
BRHA
)
t
cn
= max
_
2 ( +d
ARBR
+d
BRHA
+d
HACP
) , (18)
2 ( +d
ARBR
+d
BRCP
)
_
+2 ( +d
ARBR
+d
BRCP
)
In PMIPv6, the MT always uses the same address as it moves within the same LMA
domain. There is no latency due to movement detection or address conguration. According
to Fig. 7c, we have
T
PMIPv6
L3
= t
p
(19)
t
p
denotes the latency due to the local binding procedure, which can be expressed as
t
p
= 2 ( +d
MAGLMA
)
If the MT moves between different LMA domains, the latency due to MIPv6 handover is
also involved. In this case, T
PMIPv6
L3
is
T
PMIPv6
L3
= T
MIPv6
L3
+t
p
(20)
In SMS-MIPv6, the SMS binding procedure contributes to the L3 handover latency when
an IP session is going to be initialized. This kind of latency has already been considered in the
session initialization latency. Therefore, during the session holding time, only the IP binding
procedure needs to be considered when a handover occurs. According to Fig. 7d, we have
T
SMS-MIPv6
L3
= t
d
+t
a
+t
cn
, (21)
where t
d
and t
a
are the same as that in MIPv6. However, the return routability procedure is
not performed by SMS-MIPv6. Therefore, t
cn
is
t
cn
= 2 ( +d
ARBR
+d
BRCP
) (22)
6 Numeric Results
In this section, we evaluate the performance of SMS-MIPv6 and compare it with that of
MIPv6 and PMIPv6 with numerical results based on the analysis derived in the previous sec-
tion. The network used for evaluation is the same as that in Fig. 7a. In Fig. 7a, distance (i.e
the number of hops) between two nodes is classied in two types, i.e. intra-domain distance
and inter-domain distance. The distance between two nodes that are in the same network
domain is dened as Intra-domain distance (d
i ntra
), which is set to 3 by default. For exam-
ple, d
mt BR
(d
mt LMA
) is intra-domain distance. On the contrary, the distance between two
nodes that are in different network domains is dened as inter-domain distance (d
i nt er
), which
is set to 8 by default. For example, d
BRHA
(d
LMAHA
) , d
BRCP
(d
LMACP
) and d
HACP
are inter-domain distance. Other parameters and default values used in the evaluation are set as
follows: = 4, L
H
IPv6
= 40 Byt es, L
H
dst opt
= 20 Byt es, R = 1000 ms, D = 1, = 10 ms
and = 5.
123
232 H. Lu
(b) (a)

Fig. 9 Signaling cost in different mobility management schemes


The evaluation results show that SMS-MIPv6 achieves the best performance in terms of
signaling cost, data trafc overhead compared with MIPv6 and PMIPv6. Moreover, SMS-
MIPv6 can reduce the handover latency significantly, although it is considered as a mobility
management scheme for global mobility. However, it increases the session initialization
latency due to SMS binding through the cellular network.
Figure 9 compares the signaling cost in SMS-MIPv6 with that in MIPv6 and PMIPv6.
The interval for binding refreshing is assumed to be t
br
, i.e. t
MIPv6
br
= t
PMIPv6
br
= t
SMS-MIPv6
br
= t
br
. Let be the average number of handovers between ARs per session and be the
average number of binding refreshing per session. Then we have = E (N
h
) = / and
= 1/t
br
.
In Fig. 9a, is set to 1. That means an MT only moves across one AR during the session
holding time. In this case, no global mobility management scheme is considered for PMIPv6.
We can see that SMS-MIPv6 produces almost the same signaling cost as PMIPv6. For all
mobility management schemes, the signaling cost increases as increases. This is because
more signaling messages will be sent for binding refreshing with larger . However, the
signaling cost remains constant when is between 4.0 and 4.8. This is due to the fact that
the MT performs handover before signaling messages for binding refreshing occur.
In Fig. 9b, is set to 4. That means an MT moves across four ARs during the session
holding time. As N
R
is set to 4, the MT also moves across the border of an LMA domain.
In this case, MIPv6 must be supported for global mobility in PMIPv6, which is dened as
PMIPv6+MIPv6 in the rest of the paper. We can see that SMS-MIPv6 produce the least
signaling cost compared with PMIPv6+MIPv6 and MIPv6. For all mobility management
schemes, the signaling cost increases when increases from 1 to 4. This is because more
signaling messages will be sent for handovers with larger . The result also shows that the
signaling cost in PMIPv6+MIPv6 increases significantly compared with that in PMIPv6.
This is due to the fact that an MT must performMIPv6 for global mobility as well as PMIPv6
for local mobility. Moreover, in PMIPv6+MIPv6, signaling messages for binding refreshing
for MIPv6 are not disturbed by local mobility between ARs. On the contrary, in SMS-MIPv6
or MIPv6, these messages might be suppressed as handovers between ARs occur, which
results in constant signaling cost when 0.8 < 4.0 or 4.0 < 8.0.
Figure 10 shows the impact of the data rate ( ) and relative session size () on data trafc
overhead in different mobility schemes. The relative session size is dened as a ratio, the
session holding time to the binding time during which an MT updates location information
at its CP. It can be expressed as = 1/ (t
cn
), where t
cn
is dened as (18).
123
SMS-MIPv6: An End-to-End Mechanism 233
(a) (b)

Fig. 10 Data trafc overhead in different mobility management schemes
(b) (a)
Fig. 11 Session initialization latency in different mobility management schemes
In Fig. 10a, is set to 1,000. From (10), we can see that almost has no effect on the
data trafc overhead in MIPv6 in this case. Therefore, the data trafc overhead in SMS-
MIPv6 is almost the same as that in MIPv6, which is much less than that in PMIPv6 and
PMIPv6+MIPv6. For all mobility management schemes, the result indicates that the data
trafc overhead increases linearly as increases.
In Fig. 10b, is set to 20. It should be noted that the increase of is in a logarithm scale.
We can see that SMS-MIPv6 produces the least data trafc overhead compared with other
mobility management schemes. Moreover, has no effect on the data trafc overhead in
SMS-MIPv6 and PMIPv6. But in MIPv6 and PMIPv6+MIPv6, the data trafc overhead
decreases as increases. This is because more packets are routed by the triangular routing
with smaller . From (10), we can see that triangular routing will result in tunnel cost.
Figure 11 illustrates the impact of intra-domain distance (d
i ntra
) and inter-domain distance
(d
i nt er
) on the session initialization latency. It should be noted that the session initialization
latency is presented in a logarithm scale. We can see that the session initialization latency
in SMS-MIPv6 is much higher than that in PMIPv6 and MIPv6. In SMS-MIPv6, additional
latency (t
cell
) is introduced in SMS-MIPv6 as it should perform SMS binding through the
cellular network. t
cell
is set to 10s during the evaluation, which dominates the total session
initialization latency in SMS-MIPv6. The results also show that the session initialization
latency increases as d
i ntra
or d
i nt er
increases. This is due to the fact that packets during
session initialization have to travel more hops with larger d
i ntra
or d
i nt er
. We can see that
PMIPv6 achieves lower session initialization latency than MIPv6. In MIPv6, the triangular
123
234 H. Lu
(b) (a)
Fig. 12 Handover latency in different mobility management schemes
routing is used for packets during the session initialization, which introduces more latency
compared with the optimization routing mode.
Figure 12 shows the impact of intra-domain distance (d
i ntra
) and inter-domain distance
(d
i nt er
) on the handover latency. We can see that the handover latency in SMS-MIPv6 is
much lower than that in MIPv6. However, it is much higher than that in PMIPv6. When
global mobility between different LMA domains occurs, MIPv6 must be supported. In this
case, SMS-MIPv6 achieves much lower handover latency than PMIPv6+MIPv6. In SMS-
MIPv6, MIPv6 or PMIPv6+MIPv6, the DADprocess for address auto-conguration counts
for a large portion of the handover latency. In PMIPv6, this process is not required. Therefore,
for SMS-MIPv6, MIPv6 or PMIPv6+MIPv6, it is important to decrease the DAD latency
in order to decrease the total handover latency. The optimistic DAD [11] has been proposed
to reduce the DAD completion time.
For SMS-MIPv6, MIPv6 and PMIPv6+MIPv6, the results showthat the handover latency
increases as d
i ntra
or d
i nt er
increases. This is because signaling messages for handovers have
to travel more hops with larger d
i ntra
or d
i nt er
. However, for PMIPv6, d
i nt er
has no effect on
the handover latency. This is due to the fact that signaling messages for handovers in PMIPv6
are exchanged within an LMA domain.
7 Conclusions
In this paper, SMS-MIPv6 is proposed as an attempt that integrates mobility management in
the cellular network and the IPv6 network. It tracks the location of an MTin the IPv6 network
by hybrid binding, where the MTs current IPv6 address can be delivered through the SMS
system in the cellular network. The analysis and evaluation results show that SMS-MIPv6
achieves the best performance in terms of signaling cost, data trafc overhead compared with
MIPv6 and PMIPv6. Moreover, SMS-MIPv6 can reduce the handover latency significantly,
although it is considered as a mobility management scheme for global mobility.
However, SMS-MIPv6 increases the session initialization latency due to hybrid binding
through the cellular network. As SMS is a non-real-time and unreliable system, extension is
required for achieving better performance.
123
SMS-MIPv6: An End-to-End Mechanism 235
Acknowledgments This work is supported by National Natural Science Foundation of China (Under Grand
No.60602018, 60832005, 60736043), MOE-MS MCC Research Fund (No. 07122801), and has been per-
formed in the project of Next Generation Broadband Mobile Networks of China (2009ZX03004-002).
References
1. Lin, Y. B., & Chlamtac, I. (2001). Wireless and mobile network architectures. New York: Wiley.
2. Johnson, D., Perkins, C., & Arkko, J. (Sept 2009). Mobility Support in IPv6, draft-ietf-mext-rfc3775bis-
03.txt (work in progress), IETF.
3. Gundavelli, S. (Ed.), Leung, K., Devarapalli, V., Chowdhury, K., & Patil, B. (August 2008). Proxy
Mobile IPv6, RFC 5213, IETF.
4. Koodli, R. (Ed.). (Sept 2009). Mobile IPv6 Fast Handovers, draft-ietf-mipshop-rfc5268bis-01.txt (work
in progress), IETF.
5. Soliman, H., Castelluccia, C., ElMalki, K. & Bellier, L. (Oct 2008). Hierarchical Mobile IPv6 Mobility
Management (HMIPv6), RFC 5380, IETF.
6. Taniuchi, K., Ohba, Y., Fajardo, V., Das, S., Tauil, M., Cheng, Y., et al. (Jan 2009). IEEE 802.21: Media
Independent Handover: Features, Applicability, and Realization, IEEE Communications Magazine.
7. ETSI TS 03.485., http://www.3gpp.org/ftp/Specs/archive/03_series/03.48/.
8. Zonoozi, M. M., Dassanayake, P., & Faulkner, M. (AprMay 1996). Teletrafc modeling of cel-
lular mobile networks, In Proceedings of the IEEE vehicular technology conference (VTC96),
(pp. 12741277).
9. Han, Y., Choi, J., & Hwang, S. (2006). Reactive handover optimization in IPv6-based mobile
networks. IEEE Journal on Selected Areas in Communications, 24(9), 17581772.
10. Thomson, S., & Narten, T. (Dec 1998). IPv6 Stateless Address Autoconguration, RFC 2462, IETF.
11. Moore, N. (April 2006). Optimistic duplicate address detection, RFC 4429, IETF.
Author Biography
Hancheng Lu received the Ph.D. degree in Communication and Infor-
mation Systems from University of Science and Technology of China
(USTC), China, in 2005. Now he works as an associate professor at
Department of EEIS of USTC. His research interests include mobile
networks, vehicular networks.
123

S-ar putea să vă placă și