Yang2009

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Telecommun Syst (2009) 42: 5–15

DOI 10.1007/s11235-009-9164-y

SHIP: Cross-layer mobility management scheme based on Session


Initiation Protocol and Host Identity Protocol
Shuigen Yang · Huachun Zhou · Yajuan Qin ·
Hongke Zhang

Published online: 12 June 2009


© Springer Science+Business Media, LLC 2009

Abstract In the all-IP wireless networks, mobility manage- SIP/Mobile IP in terms of the handoff latency, the packet
ment is a crucial issue, and it can be operated at different lay- loss, and the handoff blocking probability.
ers in the traditional Transmission Control Protocol/Internet
Protocol (TCP/IP) stack. Session Initiation Protocol (SIP) Keywords Mobility management · Session initiation
is an application layer protocol used for the signaling and protocol · Host identity protocol · Cross-layer design
the mobility management without any modifications of the
lower layer protocols, especially in the voice over IP envi-
ronment. However, since it always gets the lowest priority
in the networking model, it will produce the large handoff 1 Introduction
latency, the packet loss, and the handoff blocking probabil-
ity. To provide an efficient mobility management, SIP can The fast increasing demand for wireless access to Internet
be combined with other protocols. In this paper, we pro- applications is fueled by the remarkable access of wireless
pose a cross-layer mobility management scheme based on technologies and the explosive growth of Internet. However,
SIP and the Host Identity Protocol (HIP) which is a new since the Internet was originally designed for communica-
protocol designed to provide secure and continuous com- tions between the fixed nodes, it does not support mobility
munications between two nodes by separating the identifier well. There is a strong need for some extensions to be de-
and the locator roles of the traditional IP address. In the pro- veloped to support mobility. These extensions are referred
posed scheme, the mobile node and the correspondent node
to as mobility management mechanisms. Mobility manage-
use their host identity tags to establish the session connec-
ment consists of two fundamental operations: location man-
tions, and the mobile nodes uses the HIP location update
agement and handoff management. Location management
scheme instead of the SIP location update scheme to update
refers to the task of locating a mobile node by a correspon-
its IP address when it moves to a new subnet. We discuss
dent node in order to initiate and establish a connection.
the proposed scheme’s architecture and the major mobility
procedures, including the session setup, the handoff and the Handoff management is the process by which a mobile node
location update in this paper. We also develop an analyti- keeps its connection active when it moves from one access
cal model to study the handoff performance of the proposed point to another.
scheme, the typical SIP, and the hybrid SIP/Mobile IP, re- Mobility management can be handled at different layers
spectively. The analytical results demonstrate that the pro- in the traditional Transport Control Protocol/Internet Proto-
posed scheme outperforms the typical SIP and the hybrid col (TCP/IP) stack. For example, Session Initiation Proto-
col (SIP) [1] is an application layer solution, and Mobile IP
(MIP) [2, 3] is a network layer solution. SIP was initially
S. Yang () · H. Zhou · Y. Qin · H. Zhang developed by the Internet Engineering Task Force (IETF)
National Engineering Lab on Next Generation Internet as an application layer multimedia signaling protocol. It can
Interconnection Devices, School of Electronics and Information
Engineering, Beijing Jiaotong University, 100044 Beijing, China be used to support mobility through its ability to define a
e-mail: yangsg1981@gmail.com number of specific entities and messages. However, since
6 S. Yang et al.

the application layer protocol always gets the lowest prior- also develop an analytical model to study the handoff per-
ity in the networking model, SIP will produce a long de- formance of SHIP, the typical SIP, and the hybrid SIP/MIP,
lay in the handoff. MIP provides a network layer mobility respectively.
and transparency to the higher layers. However, there are a The rest of this paper is organized as follows. Section 2
number of problems associated with M, such as the signal- gives a short overview of SIP and HIP, and some related
ing traffic overhead, the high packet loss rate, and the hand- work. Section 3 describes the proposed cross-layer mobility
off latency, thereby causing the user-perceptible deteriora- management scheme in detail. Section 4 develops an analyt-
tions of real-time traffic [4]. Although many enhanced tech- ical model of handoff performance for SHIP, SIP, and hybrid
niques can improve its performance, MIP still has the weak- SIP/MIP, respectively. Section 5 gives the numerical perfor-
ness in terms of the efficiency and the complexity. The IETF mance evaluation results. Finally, Sect. 6 concludes this pa-
also has proposed several other mobility management proto- per.
cols, like the Stream Control Transmission Protocol (SCTP)
[5] and the Network-based Localized Mobility Management
2 Background
(NetLMM) [6]. However, these protocols suffer from disad-
vantages, e.g., SCTP lacks “All-IP networks” support and In this section, we firstly provide a short overview of SIP and
NetLMM is in the nascent stages of development. HIP, which are the essential building blocks for SHIP mo-
The fundamental problem of IP mobility is the semantic bility management architecture. We then give an overview
overloading of IP address. In the traditional TCP/IP stack, of some related work.
an IP address represents not only the topological location
but also the identifier for a network interface [7]. Further- 2.1 Session Initiation Protocol
more, connections in the current Internet are identified by
the quadruple <source IP address, destination IP address, SIP is an application layer protocol used to locate the users
source port, destination port>. When a host changes its old and negotiate the parameters needed to establish the multi-
IP address to a new one, all connections bound to the host media sessions. The user is identified by the SIP Uniform
would fail since they are identified by the old, and the host Resource Identifier (URI), whose format is similar to the
has to re-establish the connections. email address format (i.e. user @ domain). The SIP user
The traditional TCP/IP stack is already heavily loaded agent registers itself with the SIP URI in the SIP registrar
down with functionalities that have been added over the server by REGISTER message, and establishes the session
years. Optimization and adding new functionalities to sup- with other user agent using INVITE message that contains
port mobility are very difficult. Currently, a new idea has session description in Session Description Protocol (SDP)
emerged for mobility by introducing a new layer. Host Iden- format. The detailed description of SIP can be found in [1].
tity Protocol (HIP) [8] is one of these new TCP/IP stacks, SIP has the potential capabilities for Internet mobility.
and it is designed by the IETF to provide secure and con- When a mobile node moves to a new subnet and obtains a
new IP address while the session is ongoing, it will send a
tinuous communications between two nodes. HIP separates
re-INVITE message with updated session description to the
the identifier and the locator roles of the traditional IP ad-
correspondent node. This re-INVITE message maintains the
dress by introducing a new host identity namespace, and a
same Call-ID of the existing session but replacing the con-
new layer between the network layer and the transport layer.
tact field of SIP header with the new IP address to inform
In HIP, the transport layer connections are no longer bound
the correspondent node where it wants to receive the future
to the IP addresses but to the host identities, and the IP ad-
messages. This re-INVITE message also replaces the c field
dresses become pure routing tags. Therefore, the change of of SDP header with the new IP address to redirect the pack-
an IP address is transparent to the transport layer connec- ets to its new location. This procedure succeeds with 200
tions, and they do not have to be broken. Hence, HIP sup- OK and ACK exchanged between the mobile node and the
ports host mobility well in nature. correspondent node. The mobile node also needs to send a
In this paper, we propose a cross-layer mobility manage- REGISTER message to the home SIP server to update the
ment scheme based on the SIP and HIP inter-working mod- location information stored there, so that the new call can
els (SHIP) to support seamless and fast handoff for the mo- be correctly redirected. Figure 1 illustrates the SIP mobility
bile node. In SHIP, the mobile node and the correspondent and its operations.
node use their host identity tags to establish the session con- SIP supports session mobility without any modifications
nections, and the mobile nodes uses the HIP location update of lower layer protocols. The IETF has adopted SIP as the
scheme instead of the SIP location update scheme to update session management for the Voice over Internet Protocol
its IP address. Therefore, SHIP is a one-suite protocol that (VoIP) streams. However, since the application layer proto-
performs seamless mobility, fast handoff, and security. We col always gets the lowest priority in the networking model,
SHIP: Cross-layer mobility management scheme based on Session Initiation Protocol and Host Identity 7

SIP will produce a long delay in the handoff. In addition, back a R1 packet that contains a puzzle to be solved by the
SIP does not support mobility in other connections that are Initiator. The idea of employing a puzzle is that the Initiator
not created under SIP currently. To break these limits, the is forced to perform a moderately expensive computation
hybrid SIP and MIP schemes are discussed to provide a full before the Responder commits its computational resources
mobility management for all applications. Some of these hy- to this connection. The R1 packet also contains the public
brid schemes can be found in Sect. 2.3. key of the Responder together with Diffie-Hellman parame-
ters. On receipt of the R1 packet, the Initiator solves the puz-
2.2 Host Identity Protocol zle and sends the responder cookie in an I2 packet together
with an IPSec Security Parameters Index (SPI) value and
HIP separates the endpoint identifier and the locator roles of its encrypted public key to the Responder. The Responder
the traditional IP address by introducing a new cryptography- verifies the solution and creates the IPSec ESP security as-
based host identity namespace, and a new host identity layer sociations. The final R2 message contains the SPI value of
between the network layer and the transport layer. Each host the Responder. After completing a HIP Base Exchange, the
has a globally unique host identity, which is the public key Initiator and Responder exchange the user traffic protected
of a public-private key pair. The host identity is often repre- with IPSec.
sented by its 128-bit long hash, the Host Identity Tag (HIT). HIP supports host mobility in nature. When the mobile
In HIP, the transport layer connections are no longer bound node moves to a new subnet and gets a new IP address, it
to the IP addresses but to the HITs, and the IP addresses performs the HIP UPDATE mechanism [11] with the corre-
become pure routing tags. spondent node. This update informs the correspondent node
In HIP, the nodes need to establish the HIP layer end-to- that it has moved to a new IP address. The transport layer
end connections using HIP Base Exchange [9]. During the connection that is running on does not have to be broken
HIP Base Exchange, a Diffie-Hellman procedure is used to because the HITs used to identify the transport layer con-
create a session key and to establish a pair of IP Security nection do not change. If the mobile node and the corre-
(IPSec) Encapsulated Security Payload (ESP) security as- spondent node change their IP addresses at roughly the same
sociations between two nodes [10]. Figure 2 illustrates the time, the HIP UPDATE messages would not be able to reach
message sequences of the HIP Base Exchange. the mobile node and the correspondent node. This double
The Initiator begins the HIP Base Exchange by sending jump problem can be solved by having the mobile node and
an I1 packet that contains both the HITs of the communica- the correspondent node update their IP addresses at the Do-
tion nodes. On receipt of the I1 packet, the Responder sends main Name System (DNS) [12] or at the Rendezvous Server
(RVS) [13].

2.3 Related work

In [14], the authors analyzed how SIP could be used to


provide an initial mapping between an HIT and an IP ad-
dress for a particular session. The authors investigated that
both HIP and SIP could provide complementary function-
ality to provide a better service behavior. However, in their
proposal, SIP and HIP worked independently to provide ser-
vices. The authors did not discuss how to use the integrated
SIP and HIP scheme for mobility management. Efficient
Fig. 1 SIP mobility and its operations
combination of HIP and SIP is still an open issue and needs
more researches.
Most of the previous work focused on the combination
of MIP and SIP. In [15], the authors used SIP for real time
and MIP for non-real time services. The authors also made
a simulation study of handoff performance based on packet
loss ratio for SIP and MIP. In [16], the authors proposed
a combined MIP and SIP mobility management approach
to provide fast handoff, and presented simulations based on
NS-2 to show performance improvement. In [17], the au-
thors integrated the MIP and SIP network entities to avoid
Fig. 2 HIP base exchange messages redundant signaling. In [18], the authors integrated SIP and
8 S. Yang et al.

MIP for macro-mobility. In [19], the authors presented a the node’s domain name and its corresponding HMS’s IP
multilayer mobility management scheme based on MIP and address.
SIP for all-IP networks where micro-mobility was handled In order to establish the sessions between two nodes, it is
separately from macro-mobility. In [20], the authors ana- necessary to have a session description format that supports
lyzed the issues related to the adoption of SIP services in the encoding of an IP address and an HIT. An extension to
the 3G networks, in particular the implications of adopting encode the HITs in the SDP is proposed in [14], which en-
SIP in the presence of MIP and the interactions between the hanced the SDP ‘k’ and ‘a’ parameters to support the encod-
two protocols within the scope of the IP Multimedia Sub- ing of an HIT. The ‘k’ parameter has the following structure:
system (IMS), focusing on IPv6. In [21], the authors pre- k=host-identity: <HIP HIT>. The ‘a’ parameter has the fol-
sented two approaches based on the hybrid MIP-SIP archi- lowing structure: a=key-mgmt: host identity <HIP HIT>.
tecture. The first approach culminated in a tightly integrated When the Initiator wants to communicate with the Re-
architecture, which merged the redundant mobility entities sponder, it firstly obtains the Responder’s HIT and the IP
in MIP and SIP to yield maximum system efficiency. The address. Then the Initiator and the Responder perform the
other approach leaded to a loosely integrated architecture, HIP Base Exchange to establish the HIP layer connection.
where their physical entities were kept intact. Once the HIP layer connection is established, the Initiator
All the above mentioned literature did not consider the and the Responder handles session establishment using the
hybrid usage of MIP or SIP based on the handoff latency, offer/answer model, which consists of a three-way negoti-
packet loss, and handoff-blocking probability. Furthermore, ation of session parameters. In the SHIP environment, the
all approaches did not consider the handoff performance due HIT replaces the IP address in the SIP header and the SDP
to the wireless link frame error rate, which plays an impor- packet. Figure 3 illustrates the communication process of
tant role in the handoff performance. SHIP.
1) Rosa sends a resolution request containing the details
of the Smith’s domain name to the DNS to obtain the
3 Cross-layer design Smith’s IP address. On receipt of this resolution request,
the DNS answers with the corresponding HMS’s IP ad-
In this section, we propose the cross-layer mobility manage- dress. These steps are illustrated by F1 and F2 in Fig. 3.
ment scheme called SHIP. We discuss the architecture and 2) After obtaining the HMS’s IP address, Rosa sends a res-
the major mobility procedures in detail, including the ses- olution request containing the details of Smith’s URI to
sion setup, the handoff and the location update. the HMS to obtain Smith’s HIT and the IP address. On
receipt of this resolution request, the HMS answers with
3.1 Architecture the HIT and the IP address of Smith. These steps are il-
lustrated by F3 and F4 in Fig. 3.
In the SHIP architecture, the functions of the SIP and HIP 3) After obtaining the HIT and the IP address of Smith,
entities are decomposed and the resulting functions of sim- Rosa sends an I1 packet of the HIP Base Exchange to
ilar functionality are integrated to minimize the redundan- Smith. Rosa and Smith start the HIP Base Exchange
cies. Architecturally, the SIP registrar server and the HIP procedures to establish the HIP layer connection. These
RVS are integrated into a uniform entity, which is referred steps are illustrated by F5, F6, F7 and F8 in Fig. 3.
to as the SHIP Home Mobility Server (HMS). The SHIP 4) Rosa sends an SIP request “INVITE” to Smith to indicate
HMS records the HITs and the IP addresses of the nodes that it wants to communicate with Smith. This request
registered at it. contains the details of Rosa’s HIT (HIT-R) and IP address
A node is identified with an HIT as the globally unique (IP-R), Smith’s HIT (HIT-S) and IP address (IP-S), ports
host identifier, and it registers its HIT and IP address at the to be used and other parameters. The SDP is used in the
HMS. A user is identified with an SIP URI as the globally payload for this purpose. This step is illustrated by F9 in
reachable user identifier, and it registers its URI at the HMS. Fig. 3.
As a result, the record in an address binding list (database) 5) Smith reads the request sent from Rosa, and tells Rosa
managed in the HMS may look like (HIT, URI, IP address). that it has received the “INVITE” message by sending
The mapping of the IP address, the HIT and the URI can a “100 Trying” message to Rosa. This step is illustrated
be established when the node initially registers at the HMS by F10 in Fig. 3.
using an extended HIP Registration Protocol [22] with its 6) While the phone rings, Smith sends provisional messages
URI enclosed, or simply by using both the HIP and the SIP (180 Ringing) to Rosa just so it does not time out and give
registration messages for this kick-off registration. When re- up. This step is illustrated by F11 in Fig. 3.
quested, the HMS will return both the HIT and the IP ad- 7) Eventually, Smith decides to accept the call. At this
dress. In addition, the DNS records the relationship between point, Smith sends a “200 OK” response to Rosa. In the
SHIP: Cross-layer mobility management scheme based on Session Initiation Protocol and Host Identity 9
Fig. 3 SHIP communication
process

payload of this response, there is another SDP message, parameter containing the MN’s current IP address, signs the
which contains a set of session parameters. All types of packet with the private key matching to the used host iden-
SIP requests are accepted using “200 OK” responses. tity, and sends the packet to the CN. This location update
This step is illustrated by F12 in Fig. 3. packet with the REA parameter is called as REA packet. The
8) On receipt of the “200 OK” response, Rosa confirms REA packet also contains an ESP_INFO parameter contain-
with an ACK message. There are no retries and no re- ing the value of the old SPI and the new SPI for the secu-
sponse messages for this request type “INVITE”, even if rity consideration. The old SPI is set to the previous SPI,
the message is lost. The ACK message is only used in and the new SPI is set to the desired new SPI value for the
the case of an INVITE message. This step is illustrated incoming security association. Optionally, the MN may in-
by F13 in Fig. 3. clude a Diffie-Hellman parameter for a new Diffie-Hellman
After these steps shown above, Rosa and Smith establish key. The MN waits for this REA packet to be acknowledged,
a session between them, which is protected by IPSec pro- and retransmits if necessary.
tocol. This session traffic is created with the HITs in SHIP, On receipt of the REA packet sent from the MN, the CN
and the header of the transport layer protocol uses the HITs starts an IP address verification process for the new IP ad-
instead of the IP addresses. dress that is included in the REA packet. The IP address
verification is needed to avoid accepting false updates. The
3.2 Mobility management CN also updates any local bindings between the HIP asso-
ciation and the MN’s IP address. The CN then sends a loca-
For a successful handoff, the Mobile Node (MN) performs a tion update packet with an Address Check (AC) parameter
location update with the HMS and the Correspondent Node to the IP address that was in the REA packet. This location
(CN) to indicate its new IP address when it moves to a new update packet with the address check parameter is called as
subnet. Figure 4 illustrates the SHIP mobility and its opera- AC packet. The AC packet also includes an ESP_INFO pa-
tions. rameter with the old SPI and the new SPI parameters which
When an MN moves to a new subnet, it obtains a new IP are set to the value of the pre-existing incoming SPI.
address from the Access Router (AR) in the new subnet. The On receipt of the AC packet that matches to the REA
MN then updates its IP address at the CN using the HIP loca- packet sent earlier, the MN responds with a location update
tion update mechanism [11]. During this procedure, the MN packet with an Address Check Reply (ACR) parameter. This
generates a location update packet with a Readdress (REA) location update packet with the address check reply parame-
10 S. Yang et al.
Fig. 4 SHIP mobility and its
operations

ter is called as ACR packet. When the CN receives the ACR and the wired link between the access router and the CN. We
packet, it uses the new IP address as the location information define the following parameters at first.
of the MN and removes the old IP address. D One-way wireless link delay between the MN and
Since the header of the transport layer protocol uses the the access router.
HITs instead of the IP address, the MN does not have to send tac One-way wired link delay between the access router
the re-INVITE message to the CN. Therefore, SHIP shows and the CN.
no difference for the transport layer protocol even though tmc One-way packet transportation delay between the
the IP address is changed. MN and the CN.
In order to be established a session by a new CN, the The entire handoff latency Th includes the time required
MN should update its location at the HMS, though it does for the MN’s link layer handoff to the new subnet tl2 , the
not factor into the handoff delay. The location update proce- time required for the new IP address acquisition by the MN
dures at the HMS are almost the same as the location update in the new subnet ta , and the time required for the location
procedures at the CN without including the ESP_INFO pa- update to the CN tup . Therefore,
rameter. The MN generates a REA packet that contains the
current IP address of the MN, and sends the packet to the Th = tl2 + ta + tup . (1)
HMS. When the HMS receives the REA packet, it updates
According to the SIP mobility and its operations shown
the address binding list between the MN’s HIT and the IP
in Fig. 1, the location update to the CN includes the re-
address, and then sends an AC packet to the MN. On receipt
INVITE, the OK, and the ACK procedures. According to
of the AC packet sent from the HMS, the MN responds with
the SHIP mobility and its operations shown in Fig. 4, the lo-
an ACR packet.
cation update to the CN includes the Readdress (REA), the
Address Check (AC), and the Address Check Reply (ACR)
procedures. Therefore, the location update to the CN in both
4 Analytical models for handoff performance
SHIP and SIP includes three procedures. On the other hand,
in the hybrid SIP/MIP (MSIP), since the MN and the CN
In this section, we develop an analytical model to inves-
need to use the home agent to re-direct the data packets, the
tigate the handoff performance for SHIP, the typical SIP,
MN has to update its IP address at the CN and the home
and the hybrid SIP/MIP, respectively. In all-IP wireless net-
agent [16, 21]. Each location update includes two proce-
works, the handoff performance can be specified in terms of
dures, binding update request and acknowledgement. As-
the handoff latency, the packet loss, and the handoff block-
sume that the one-way wired link delay between the access
ing probability. Analyses of these metrics are very useful in
router and the home agent is equal to the one-way wired link
evaluating the performance of mobility management proto-
delay between the access router and the CN. Therefore,
cols. For simplicity, we call the hybrid SIP/MIP as MSIP.

3tmc , for SHIP and SIP,
4.1 Handoff latency tup = (2)
4tmc , for MSIP.

The path between the MN and the CN contains two parts: Let pf denote the wireless link layer frame error rate, τ
the wireless link connecting the MN and the access router, denote the wireless link layer inter-frame interval, and pij
SHIP: Cross-layer mobility management scheme based on Session Initiation Protocol and Host Identity 11

denote the probability that the first frame transmitted by the the likelihood that a session/call connection will be prema-
MN is received correctly by the access router, being the ith turely terminated due to unsuccessful handoff during a ses-
retransmitted frame at the j th retransmission trial. The ex- sion lifetime. The handoff blocking probability is consid-
pression for pij and the one-way frame transportation delay ered to be more important than the blocking probability of
tf between the MN and the access router with radio link new calls because the call is already active and the quality
protocol are given by [23] of service is more sensitive for the handoff calls. Hence, the
minimization of the handoff blocking probability is crucial
i 2 −i
pij = pf (1 − pf )2 [(2 − pf )pf ]( 2 +j −1) , (3) for mobility management schemes.
The handoff blocking can be caused by many factors,

n 
i such as the handoff latency, the signal-to-noise deterioration,
tf = D(1 − pf ) + pij [2i · D + 2(j − 1)τ ], (4) the unavailable channel, and the session rejection by the net-
i=1 j =1 work. However, we only consider the handoff latency as the
handoff blocking factor in this paper. In this scenario, if the
where i = 1, 2, . . . , n, j = 1, 2, . . . , i, and n is the maximum
subnet residence time is less than the total handoff time, the
number of trials that the radio link protocol carries out be-
service is forcefully terminated due to the loss of the link or
fore aborting the attempt to transmit a frame over the link
the channel.
layer. Typical, n = 3.
Let Th denote the total handoff latency, and its mean
Let Lp and Lf denote the length of a packet (typically
value is E[Th ]. Let tc denote the subnet residence time, and
an IP packet) and the length of a wireless link layer frame,
its probability density function is fc (t). Let μc denote the
respectively. Therefore, the number of frames per packet
L border crossing rate for the MN out of a subnet. For the sake
is K =  Lfp . The one-way packet transportation delay be-
of simplicity and to easily derive the analytical expression,
tween the MN and the CN tmc is we assume that Th is exponentially distributed with the cu-
mulative function FT (t). Then, the handoff blocking proba-
tmc = tf + (K − 1)τ + tac . (5)
bility (PB ) is given by
By summing up (1) and (2), we would get the handoff
PB = Pr (Th > Tc )
latency as follows.  ∞
 = [1 − FT (u)]fc (u)du
t + t + 3tmc , for SHIP and SIP,
Th = l2 a (6) 0
tl2 + ta + 4tmc , for MSIP, μc E[Th ]
= . (8)
where tmc is given by (5). 1 + μc E[Th ]

According to [24], if we assume that the access router’s


4.2 Packet loss coverage area is circular, the border crossing rate for the MN
out of a subnet is given by
Without efficient buffer management, the packets transmit-
ted by the CN during the handoff latency will be lost. There- 2ν
μc = (9)
fore, if the packet transmission rate of the CN is λ, the num- πR
ber of packets that will be lost during the handoff (PL ) is
where ν is the average velocity of the MN, and R is the
given by
radius of the subnet.

λ(tl2 + ta + 3tmc ), for SHIP and SIP,
PL = λ · Th (7)
λ(tl2 + ta + 4tmc ), for MSIP.
5 Performance evaluation results
According to (7), we can see that the packet loss during
a handoff is proportional to the handoff latency. In this section, we give the performance evaluation results
of the handoff performance for SHIP, SIP, and MSIP, re-
4.3 Handoff blocking probability spectively. For the numerical calculations, we use the fol-
lowing parameter values used in the previous work [23–25]:
Handoff blocking probability represents the probability that tac = 50 ms, tl2 = 10 ms, ta = 20 ms, n = 3, τ = 20 ms, and
a handoff request is blocked due to the lack of available Lf = 19 bytes. The wireless link delay between the MN and
channels. It was computed as the ratio of the total number the access router is D = 10 ms and 150 ms for the WLAN
of blocked handoffs to the total number of handoff requests and the satellite network, respectively [26]. The lengths of
during a simulation period. And it is often used to express the SIP re-INVITE message, the SHIP UPDATE message,
12 S. Yang et al.

Fig. 5 Handoff latency versus wireless link layer frame error rate Fig. 6 Packet loss versus wireless link layer frame error rate

and the MSIP binding update message are 140 bytes [25], Assume the packets transmission rate of the CN is λ = 10
80 bytes [9, 11], and 56 bytes [16, 21], respectively. packets/sec [24]. Let the wireless link layer frame error rate
Let the wireless link layer frame error rate varies from 0 varies from 0 to 0.5 with a step value of 0.05. Figure 6 shows
to 0.5 with a step value of 0.05. Figure 5 shows the handoff the packet loss against the wireless link layer frame error
latency against the wireless link layer frame error rate. rate. The packet loss increases as the wireless link layer
We can get two conclusions from Fig. 5: frame error rate increases, and the packet loss of SHIP is
1) The handoff latencies of SHIP, SIP, and MSIP increase as lower than that of SIP and MSIP. In fact, the packet loss
the wireless link layer frame error rate increases. This can during a handoff is proportional to the handoff latency in the
be explained as follows. A higher value of the wireless proposed analytical model. Therefore, the packet loss has a
link layer frame error rate increases the probability of the similar nature to the handoff latency.
erroneous packet transfer across the wireless link. Since Let the average velocity of the MN is ν = 20 m/s, the ra-
the handoff signaling messages have to be retransmitted dius of the subnet is R = 500 m, and the wireless link layer
by the MN several times before the access router receives frame error rate varies from 0 to 0.5 with a step value of 0.05.
them, it increases the wireless link layer packet transfer Figure 7 shows the handoff blocking probability against the
delay and then results in higher handoff latency. More- wireless link layer frame error rate. The handoff blocking
over, the handoff latency increases faster as the wireless probability increases as the wireless link layer frame error
link layer frame error rate increases when the wireless rate increases, and the handoff blocking probability of SHIP
link layer access delay is larger. is lower than that of SIP and MSIP. The proposed analyti-
2) The handoff latency of SHIP is lower than that of SIP and cal model only considers the handoff latency as the handoff
MSIP. This can be explained as follows. The transmis- blocking probability factor in the performance evaluation.
sion delay across the wireless link depends on the num- Therefore, if we fixed the average velocity of the MN and
ber of the wireless link layer frames. The larger size of the radius of the subnet, the handoff blocking probability
the packet means the higher probability that it gets erro- has a similar nature to the handoff latency.
neous packet during its transfer across the wireless link Let the wireless link layer frame error rate is pf = 0.2,
layer. This increases the wireless link layer packet trans- the radius of the subnet is R = 500 m, and the average ve-
fer delay and then results in higher handoff latency. As locity of the MN varies from 0 to 40 m/s with a step value
the size of an SIP handoff signaling message is larger of 5 m/s. Figure 8 shows the handoff blocking probability
than that of an SHIP handoff signaling message, the SIP against the average velocity of the MN. The handoff block-
handoff signaling message requires a higher number of ing probability increases as the average velocity of the MN
the retransmissions, and then the higher handoff latency. increases. A higher value of the average velocity of the MN
On the other hand, although the size of an MSIP handoff increases the border crossing rate for the MN out of a sub-
signaling message is smaller than that of an SHIP handoff net. Therefore, the MN has to update its IP address more
signaling message, the MN has to update its IP address at frequently. This results in the higher handoff latency and the
both the CN and the home agent. Therefore, the handoff handoff blocking probability. Moreover, the handoff block-
latency of MSIP will be higher than that of SHIP. ing probability increases faster as the average velocity of
SHIP: Cross-layer mobility management scheme based on Session Initiation Protocol and Host Identity 13

Fig. 7 Handoff blocking probability versus wireless link layer frame Fig. 9 Handoff blocking probability versus radius of the subnet
error rate

Fig. 9, we can also see that the handoff blocking probability


of SHIP is lower than that of SIP and MSIP. This is due to
the smaller size of an SHIP handoff signaling message than
that of SIP, and the less location update procedures in SHIP
than in MSIP.
Based on the numerical handoff performance investiga-
tion shown above, we can conclude that the proposed SHIP
outperforms SIP and MSIP in terms of reducing the handoff
latency, the packet loss, and the handoff blocking probability
under various scenarios.

6 Conclusion

Due to the historical reasons, mobility management in the


all-IP wireless networks is crucial. SIP is widely used for the
signaling and the mobility management without any modifi-
Fig. 8 Handoff blocking probability versus average velocity of the
cations of the lower layer protocols, but it always produces
mobile node
the large handoff latency and the packet loss. To enable a
better handoff performance, we propose a cross-layer mo-
the MN increases when the wireless link layer access delay bility management scheme based on SIP and HIP. The basic
is larger. In addition, from Fig. 8, we can also see that the idea of the proposed scheme is that the mobile node and
handoff blocking probability of SHIP is lower than that of the correspondent node use their host identity tags to es-
SIP and MSIP. This is due to the smaller size of an SHIP tablish the session connections, and the mobile nodes uses
handoff signaling message than that of SIP, and the less lo- the HIP location update scheme instead of the SIP location
cation update procedures in SHIP than in MSIP. update scheme to update its location. Therefore, the pro-
Let the wireless link layer frame error rate is pf = 0.2, posed scheme has a one-suite protocol stack that performs
the average velocity of the MN is ν = 20 m/s, and the ra- the seamless mobility and the fast handoff. The analytical
dius of the subnet R varies from 400 m to 800 m with a results demonstrate that the proposed scheme outperforms
step value of 50 m. Figure 9 shows the handoff blocking the typical SIP and the hybrid SIP/Mobile IP in terms of the
probability against the radius of the subnet. As shown in handoff latency, the packet loss, and the handoff blocking
Fig. 9, the handoff blocking probability decreases as the ra- probability.
dius of the subnet increases. A higher value of the radius In the proposed scheme, when two nodes want to com-
of the subnet decreases the border crossing rate for the MN municate with each other, they firstly need to establish the
out of a subnet. Therefore, it results in the lower handoff la- HIP layer connection, and then the session connection us-
tency, and then the lower handoff blocking probability. From ing the host identity tags. That is, the session connection is
14 S. Yang et al.

based on the HIP layer connection. As a result, the proposed 16. Jung, J., Mudumbai, R., Montgomery, D., & Kahng, H. (2003).
scheme needs more initial setup time than the typical SIP- Performance evaluation of two layered mobility management us-
ing mobile IP and session initiation protocol. In Proceedings
based mobility management scheme.
of the IEEE Global Telecommunications Conference (GLOBE-
Work in progress is to validate the analytical numerical COM’2003) (vol. 3, pp. 1190–1194), San Francisco, USA.
results by the extensive simulations using the OMNeT++, 17. Wang, Q., Abu-Rgheff, M. A., & Akram, A. (2004). Design and
which is a public source, modular, component-based and evaluation of an integrated mobile IP and SIP framework for ad-
vanced handoff management. In The 2004 IEEE International
open-architecture discrete event simulation system. It is also
Conference on Communications (ICC’2004) (pp. 3921–3925),
very interesting to extend the proposed cross-layer scheme Paris, France.
to other mobility types beyond the terminal handoffs. 18. Wong, K. D., Dutta, A., Burns, J., Jain, R., & Young, K. (2003).
A multilayered mobility management scheme for auto-configured
Acknowledgement We would like to thank the anonymous review- wireless IP networks. IEEE Wireless Communications, 10(5), 62–
ers for their detailed reviews, helpful suggestions and constructive 69.
comments, which have helped to improve the quality of this paper. 19. Politis, C., Chew, K. A., Akhtar, N., Georgiades, M., & Tafa-
This work is supported in part by the National Basic Research Pro- zolli, R. (2004). Hybrid multilayer mobility management with
gram of China (“973 program”) under contract No. 2007CB307101 AAA context transfer capabilities for all-IP networks. IEEE Wire-
and No. 2007CB307106, in part by the Cultivation Fund of the Key less Communications, 11(4), 76–88.
Scientific and Technical Innovation Project, Ministry of Education of 20. Faccin, S. M., Lalwaney, P., & Patil, B. (2004). IP multimedia ser-
China under contract No. 706005, and in part by the National Natural vices: analysis of mobile IP and SIP interactions in 3G networks.
Science Foundation of China under contract No. 60870015. IEEE Communications Magazine, 42(1), 113–120.
21. Wang, Q., & Abu-Rgheff, M. A. (2006). Mobility management
architectures based on joint mobile IP and SIP protocols. IEEE
Wireless Communications, 13(6), 68–76.
References 22. Laganier, J., Koponen, T., & Eggert, L. (2006). Host identity pro-
tocol (HIP) registration extension. IETF Internet Draft, draft-ietf-
1. Handley, M., Schulzrinne, H., Schooler, E., & Rosenberg, J. hip-registration-02.
(1999). SIP: session initiation protocol. IETF RFC 2543. 23. Bao, G. (1996). Performance evaluation of TCP/RLP protocol
2. Perkins, C. E., Patil, B., & Roberts, P. (2002). IP mobility support stack over CDMA wireless link. Journal of Wireless Networks
for IPv4. IETF RFC 3220. (WINET), 2(3), 229–237.
3. Johnson, D. B., Perkins, C. E., & Arkko, J. (2004). Mobility sup- 24. Makaya, C., & Pierre, S. (2008). An analytical framework for per-
port in IPv6. IETF RFC 3775. formance evaluation of IPv6-based mobility management proto-
4. Le, D., Fu, X., & Hogrefe, D. (2006). A review of mobility sup- cols. IEEE Transactions on Wireless Communications, 7(3), 972–
port paradigms for the Internet. IEEE Communications Survey & 983.
Tutorials, 8(1), 38–51. 25. Mohanty, S., & Akyildiz, I. F. (2007). Performance analysis of
5. Stewart, R., Xie, Q., Morneault, K., Sharp, C., Schwarzbauer, H., handoff techniques based on mobile IP, TCP-migrate, and SIP.
Taylor, T., Rytina, I., Kalla, M., Zhang, L., & Paxson, V. (2000). IEEE Transactions on Mobile Computing, 6(7), 731–747.
Stream control transmission protocol. IETF RFC 2960. 26. Akan, O. B., & Akyildiz, I. F. (2004). ATL: an adaptive transport
6. Kempf, J. (2007). Goals for network-based localized mobility layer suite for next-generation wireless Internet. IEEE Journal on
management. IETF RFC 4831. Selected Areas in Communications, 22(5), 802–817.
7. Balakrishnan, H., Lakshminarayanan, K., Ratnasamy, S., Shenker,
S., Stoica, I., & Walfish, M. (2004). A layered naming architecture
for the Internet. ACM SIGCOMM Computer Communication Re-
view, 34(4), 343–352. Shuigen Yang received his B.S.
8. Moskowitz, R., & Nikander, P. (2006). Host identity protocol degree from Beijing Jiaotong Uni-
(HIP) architecture. IETF RFC 4423. versity (formerly knows as North-
9. Moskowitz, R., Nikander, P., Jokela, P., & Henderson, T. R.
ern Jiaotong University), Beijing,
(2007). Host identity protocol. IETF Internet Draft, draft-ietf-hip-
China, in June 2004. He is currently
base-09.
working toward the Ph.D. degree
10. Jokela, P., Moskowitz, R., & Nikander, P. (2007). Using ESP trans-
at the Next Generation Internet Re-
port format with HIP. IETF Internet Draft, draft-ietf-hip-esp-06.
search Center, School of Electronics
11. Henderson, T. (2007). End-host mobility and multihoming with
and Information engineering, Bei-
the host identity protocol. IETF Internet Draft, draft-ietf-hip-mm-
jing Jiaotong University, Beijing,
05.
China. His research interests include
12. Nikander, P., & Laganier, J. (2007). Host identity protocol (HIP)
the pervasive networking, and the
domain name system (DNS) extensions. IETF Internet Draft,
IP mobility in wireless environment,
draft-ietf-hip-mm-dns-09.
especially the mobility management
13. Laganier, J., & Eggert, L. (2006). Host identity protocol (HIP)
in the wireless Internet.
rendezvous extension. IETF Internet Draft, draft-ietf-hip-rvs-05.
14. Tschofenig, H., Ott, J., Schulzrinne, H., Henderson, T. R., & Ca-
marillo, G. (2007). Interaction between SIP and HIP. IETF Internet
Draft, draft-tschofenig-hiprg-host-identities-05.
15. Politis, C., Chew, K. A., & Tafazolli, R. (2003). Multilayer
mobility management for all-IP networks: pure SIP vs. hybrid
SIP/mobile IP. The 57th IEEE Semiannual Vehicular Technology
Conference (VTC’2003-Spring) (vol. 4, pp. 2500–2504), Jeju, Ko-
rea.
SHIP: Cross-layer mobility management scheme based on Session Initiation Protocol and Host Identity 15
Huachun Zhou received his B.S. Huawei. She is also the chief researcher of 5 research projects founded
degree from People’s Police Officer by the Chinese government. Her research interests are in the areas of
University of China in 1986. He re- computer networks and wireless communications.
ceived his M.S.E. degree from Bei-
jing Jiaotong University (formerly
knows as Northern Jiaotong Univer- Hongke Zhang received his M.S.
sity), Beijing, China, in 1989. His and Ph.D. degrees in Electrical and
research interests include the com- Communication Systems from the
puter network and the network man- University of Electronic Science
agement technologies. and Technology of China (formerly
known as Chengdu Institute of Ra-
dio Engineering) in 1988 and 1992,
respectively. From Sep. 1992 to
June 1994, he was a post-doc re-
search associate at Beijing Jiao-
tong University (formerly known as
Northern Jiaotong University). In
Yajuan Qin received her B.S. and July 1994, he jointed the School of
M.S. degrees in Electrical Engineer- Electronics and Information Engi-
ing from the University of Elec- neering, Beijing Jiaotong Univer-
tronic Science and Technology of sity, where he is now a professor. He is also the dean of the School of
China (formerly known as Chengdu Electronics and Information Engineering and the director of the Next
Institute of Radio Engineering) in Generation Internet Research Center. He has published more than 100
1985 and 1988, respectively, and research papers in the areas of communications, computer networks
Ph.D. degree in communication en- and information theory. He is the author of eight books written in Chi-
gineering from Beijing University nese. He was the chief investigator of more than 20 research projects
of Posts and Telecommunications founded by Chinese government and corporations such as Intel, cisco,
in 2003. From Jan. 2002 to April Huawei. He is now the chief scientist of the projects “a universal net-
2002, she was a research associate work architecture for supporting pervasive services” founded by the
at CRL, Japan. In 2003, she joined National Basic Research Program of China (“973 Program”). He is
Beijing Jiaotong University (for- also a member of the electronics and information science steering com-
merly known as Northern Jiaotong mittee of the Ministry of Education, a member of the expert committee
University), where she is now an associate professor. She has pub- of the national Natural Science Foundation of China (NSFC), and a
lished more than 30 research papers and is the holder of more than member of the expert committee of the Ministry of Information Indus-
ten patents. She is also the principle researcher of about ten research try.
projects founded by the Chinese Government and corporations such as

You might also like