3, MARCH 2013
1077
AbstractIPv6 mobility management is one of the most challenging research topics for enabling mobility service in the forthcoming mobile wireless ecosystems. The Internet Engineering
Task Force has been working for developing efficient IPv6 mobility
management protocols. As a result, Mobile IPv6 and its extensions
such as Fast Mobile IPv6 and Hierarchical Mobile IPv6 have been
developed as host-based mobility management protocols. While
the host-based mobility management protocols were being enhanced, the network-based mobility management protocols such
as Proxy Mobile IPv6 (PMIPv6) and Fast Proxy Mobile IPv6
(FPMIPv6) have been standardized. In this paper, we analyze and
compare existing IPv6 mobility management protocols including
the recently standardized PMIPv6 and FPMIPv6. We identify
each IPv6 mobility management protocols characteristics and
performance indicators by examining handover operations. Then,
we analyze the performance of the IPv6 mobility management
protocols in terms of handover latency, handover blocking probability, and packet loss. Through the conducted numerical results,
we summarize considerations for handover performance.
Index TermsFast Mobile IPv6 (FMIPv6), Fast Proxy Mobile
IPv6 (FPMIPv6), Hierarchical Mobile IPv6 (HMIPv6), Mobile
IPv6 (MIPv6), Proxy Mobile IPv6 (PMIPv6).
I. I NTRODUCTION
OBILE wireless ecosystems facilitate more rapid
growth of digital ecosystems for our human lives
[1][6]. Mobility management protocols are at the heart of the
mobile wireless ecosystems. Mobile social networking, mobile
collaboration computing, and mobile shopping shall become a
reality with a well-deployed mobility management architecture.
Various mobility management protocols for enabling mobility service have been introduced. In particular, mobility
support in the network layer has been being developed by the
Internet Engineering Task Force (IETF). Since the Mobile IPv6
(MIPv6) specification [7] was published, extensions including
Fast Mobile IPv6 (FMIPv6) [8] and Hierarchical Mobile IPv6
(HMIPv6) [9] for enhancing the performance of MIPv6 have
been developed. During the time when the extensions to MIPv6
1078
and HMIPv6 have been compared and evaluated in terms of signaling cost, binding refresh cost, packet delivery cost, required
buffer space, and handover latency. In the paper, the authors
presented the effect of subnet residence time, packet arrival
rate, and wireless link delay to the different IPv6 mobility
management protocols.
Simple handover performance analysis has been presented
in [16]. In the paper, the authors showed that PMIPv6 outperforms other IPv6 mobility management protocols owing to
its simple handover procedure. In [17], HMIPv6 and PMIPv6
are compared and analyzed in terms of location update, packet
delivery, and wireless power consumption costs. Then, in [18],
four different route optimization (RO) schemes for PMIPv6
are presented and analyzed. In the paper, the authors have
showed that the router optimization schemes solve the ineffective routing path problem and argued that the scalability of the
PMIPv6 architecture is improved owing to distributed routing
paths in the router optimization schemes. In [19], an analytical
cost model has been developed for evaluating the performance
of IPv6 mobility management protocols. The IPv6 mobility
management protocols such as MIPv6, FMIPv6, HMIPv6, and
PMIPv6 are analyzed and compared in terms of signaling cost,
packet delivery cost, tunneling cost, and total cost.
However, the previous performance analysis studies [12],
[13] considered only the host-based mobility management protocols. In [16] and [19], PMIPv6 has been compared with the
host-based mobility management protocols, but the recently
developed FPMIPv6 protocol [15] has not been considered.
Moreover, the cost analysis studies performed in [13] and
[17][19] do not help to understand the handover performance
of IPv6 mobility management protocols.
In this paper, we develop a uniform framework for conducting analytic modeling across the spectrum of IPv6 mobility
management protocols. The host-based mobility management protocols such as MIPv6, FMIPv6, and HMIPv6 and
the network-based mobility management protocols such as
PMIPv6 and FPMIPv6 are analyzed and compared in terms
of handover latency, handover blocking probability, and
packet loss.
III. P RELIMINARIES
A. Performance Metrics
The following performance metrics are used.
1) Handover latency: It is the time interval during which an
MN cannot send or receive any packets while it performs
its handover between different access networks.
2) Handover blocking probability: It is the probability which
an MN cannot complete its handover when the network
residence time is less than the handover latency.
3) Packet loss: It is the sum of all lost packets destined for
an MN during the MNs handover.
B. Considered Network Model
The considered network model is depicted in Fig. 1 showing a generic network topology wherein all communication
entities are displayed. Suppose that the MN changes its point
LEE et al.: COMPARATIVE HANDOVER PERFORMANCE ANALYSIS OF IPv6 MOBILITY MANAGEMENT PROTOCOLS
Fig. 1.
1079
n
i
i=1 j=1
1080
i)/2)+j1)
(2)
(4)
(5)
where BWwired and Dwired are the bandwidth and the latency
of wired links, respectively. Then, by considering the number
of hops between the two end nodes, the one-way packet transportation delay over a number of wired links dwd (Lp , h) is
obtained as follows [26]:
dwd (Lp , h) =
Lp h
+ Dwired
BWwired
(6)
LHO
(7)
(8)
LEE et al.: COMPARATIVE HANDOVER PERFORMANCE ANALYSIS OF IPv6 MOBILITY MANAGEMENT PROTOCOLS
Fig. 3.
Fig. 4.
(9)
(10)
(11)
(12)
(13)
1081
LHO
= TL2 + TPRE
(14)
where TPRE represents the time at which the nAR receives the
UNA message sent from the MN and the time at which the MN
receives the first data packet sent from the nAR. Note that the
data packets sent to the MN are the buffered data packets that
the pAR has forwarded. Then, TPRE is expressed as follows:
TPRE = dwl (LUNA ) + dwl (LD ).
(15)
Fig. 4 shows the timing diagram for reactive FMIPv6 handover. Even if an MN can anticipate its movement by utilizing
the L2 trigger, sometime, the MN cannot complete its handover
preparing at the previous access network. That is, reactive
FMIPv6 handover is performed when the MN cannot receive
the FBAck message sent from the pAR [10].
1082
LHO
(16)
(17)
where dbuff -packet is the time which the first data packet
buffered at the pAR arrives at the MN via the nAR. The buffered
data packets at the pAR are immediately sent to the nAR with
the FBAck message. Accordingly, dbuff -packet is expressed as
dbuff -packet = dwd (LD + LT , hAA ) + dwl (LD )
(18)
(HMIPv6)
LHO
(19)
(20)
where dmap-packet is the time which the first data packet sent
from the MAP arrives at the MN. The MAP immediately sends
data packets destined for the MN with the LBAck message.
Accordingly, dmap-packet is expressed as
dmap-packet = dwl (LD + LT ) + dwd (LD + LT , hGA ) (21)
where LT is taken into account because the data packets sent
from the MAP to the MN are tunneled.
D. Handover Latency of PMIPv6
Fig. 6 shows the timing diagram for PMIPv6 handover.
Similar to HMIPv6, PMIPv6 manages the movement of an
MN in a localized manner as well, but mobility service for
the MN is supported by mobility service provisioning entities
[17], [18]. As the MN attaches to the new access network, its
movement is detected and registered by the MAG at the new
access network. Then, the MN obtains the same HNP included
in the RA message sent from the MAG at the new access
network so that the address configuration and DAD process are
not required when the MN performs its handover in a PMIPv6
domain [16].
LEE et al.: COMPARATIVE HANDOVER PERFORMANCE ANALYSIS OF IPv6 MOBILITY MANAGEMENT PROTOCOLS
Fig. 7.
Fig. 8.
LHO
= TL2 + TLMA
(22)
(24)
where LT is only taken into account at dwd (Lp , h). This is because the data packets for the MN are only tunneled between the
LMA and the MAG. Notice that this is a difference compared to
that of HMIPv6. Even if both of PMIPv6 and HMIPv6 similarly
manage the MN in a localized manner, PMIPv6 further reduces
the packet transportation overhead over the wireless link [17].
E. Handover Latency of FPMIPv6
Similar to FMIPv6, FPMIPv6 consists of predictive and
reactive modes.
Fig. 7 shows the timing diagram for predictive FPMIPv6 handover. While an MN is attached to a previous MAG (pMAG),
it reports an imminent handover event to the pMAG. Pre-
1083
LHO
= TL2 + TPRE-P
(25)
where TPRE-P is composed of the sum of the IP-layer connection setup delay D and the first data packet arrival delay
from the nMAG to the MN dmag-packet . Accordingly, TPRE-P
is expressed as follows:
TPRE-P = D + dmag-packet
(26)
LHO
= TL2 + TRE-P
(27)
1084
from the pMAG to the nMAG and then sent to the MN. That
is, TRE-P is expressed as follows:
TRE-P = D + dwd (LHI , hAA ) + dwd (LHAck , hAA )
+ dbuff -packet .
(28)
()
LHO
> TR =
0
()
1 FT (u) fR (u)du
()
c E LHO
=
()
1 + c E LHO
(29)
2
R
(30)
()
p = s E(S)LHO
(31)
Fig. 9.
loss will not occur owing to packet buffering facilities, but only
delayed packet transportation will occur [13].
V. N UMERICAL A NALYSIS R ESULTS AND D ISCUSSIONS
In this section, the performance evaluation results of the
mobility management protocols are presented. For the numerical analysis, the following system parameter values are used
[25][27], [31]: hCH = 4, hCG = 6, hHG = 4, hGA =
4, hAM = 1, E(S) = 10, = 20 ms, n = 3, Lf = 19 B,
Dwl = [10, 40] ms, Dwired = 0.5 ms, BWwired = 100 Mbps,
TL2 = 45.35 ms, and TDAD = 1000 ms.
A. Handover Latency
Let f vary from 0 to 0.7 with a step value of 0.05. Figs. 9 and
10 show the handover latency against f . A higher value of f
increases the probability of the erroneous packet transmission
over the wireless link. Accordingly, the number of mobility signaling retransmissions is increased, which results in increased
handover latency. In other words, as shown in Figs. 9 and 10,
the handover latency for each mobility management protocol is
relative to f . The value of Dwl also contributes to the handover
latency. For instance, the handover latency is dramatically
increased as the value of f is increased with a higher value
of Dwl . Predictive FMIPv6 and FPMIPv6 outperform the other
mobility management protocols in terms of handover latency
in this analysis. This is because an MN in those predictive
fast handover protocols utilizes the L2 trigger and prepares
its handover at the previous (current) access network before it
actually moves to the new access network. However, reactive
fast handover protocols cannot significantly reduce the handover latency because an MN in those protocols must perform
some actions at the new access network. Accordingly, from
these results, it is confirmed that the reactive fast handover
protocols such as reactive FMIPv6 and FPMIPv6 can be used to
prevent packet loss but not to significantly reduce the handover
latency. Then, PMIPv6 is placed second in this analysis. An
MN in PMIPv6 is locally managed, and mobility signaling is
LEE et al.: COMPARATIVE HANDOVER PERFORMANCE ANALYSIS OF IPv6 MOBILITY MANAGEMENT PROTOCOLS
1085
1086
Fig. 15.
VI. C ONCLUSION
In this paper, the existing IPv6 mobility management protocols developed by the IETF have been analyzed and compared
in terms of handover latency, handover blocking probability,
and packet loss. From the conducted analysis results, the following are confirmed.
1) Utilizing L2 information: In order to improve the handover performance, L2 information should be utilized.
As shown in Fig. 10, predictive FMIPv6 and FPMIPv6
outperform the other mobility management protocols because those protocols allow an MN to prepare its handover before the MN performs its actual handover to the
new access network. The reduced handover latency also
results in the reduced handover blocking probability as
shown in Figs. 1113.
2) Employing buffering management: In order to prevent
packet loss during the handover, any buffering mechanism should be employed. As shown in Figs. 14 and
15, only fast handover protocols such as FMIPv6 and
FPMIPv6 prevent the loss of data packets sent from
the CN.
LEE et al.: COMPARATIVE HANDOVER PERFORMANCE ANALYSIS OF IPv6 MOBILITY MANAGEMENT PROTOCOLS
1087
1088
Tai-Myoung Chung (SM00) received the B.S. degree in electrical engineering from Yonsei University, Seoul, Korea, in 1981, the B.S. degree in
computer science and the M.S. degree in computer
engineering from the University of Illinois, Chicago,
in 1984 and 1987, respectively, and the Ph.D. degree
in computer engineering from Purdue University,
West Lafayette, IN, in 1995.
He is currently a Professor with Sungkyunkwan
University, Suwon, Korea. His research interests are
information security, information management, and
protocols in next-generation networks.
Dr. Chung is currently the Vice-Chair of the Organisation for Economic
Co-operation and Development Working Party on Information Security and
Privacy. He serves as a Presidential Committee member of the Korean
e-government and the Chair of the Information Resource Management Committee of the e-government. He is also an expert member of the Presidential
Advisory Committee on Science and Technology of Korea and is the Chair of
the Consortium of Computer Emergency Response Teams.