Developing route optimization based pmi pv6 testbed for reliable packet transmission

Received February 18, 2016, accepted March 7, 2016, date of current version March 23, 2016.
Digital Object Identifier 10.1109/ACCESS.2016.2541164
Developing Route Optimization-Based PMIPv6
Testbed for Reliable Packet Transmission
BYUNGSEOK KANG1, NAMYEONG KWON2, AND HYUNSEUNG CHOO1, (Member, IEEE)
1Department of Computer Science and Engineering, Sungkyunkwan University, Suwon 440-746, South Korea
2Samsung Electronics, Suwon 16530, South Korea
Corresponding author: H. Choo (choo@skku.edu)
This work was supported in part by the Ministry of Science, ICT and Future Planning, South Korea, Institute for Information and
Communications Technology Promotion through the G-ITRC Program under Grant IITP-2015-R6812-15-0001 and in part by
the National Research Foundation of Korea within the Ministry of Education, Science and Technology
through the Priority Research Centers Program under Grant 2010-0020210.
ABSTRACT Proxy Mobile IPv6 (PMIPv6) allows a mobile node to communicate directly to its peers
while changing the currently used IP address. This mode of operation is called route optimization (RO).
In the RO process, the peer node learns a binding between the home address and its current temporary
care-of-address. Many schemes have been proposed to support RO in PMIPv6. However, these schemes
do not consider the out-of-sequence problem, which may happen between the existing path and the newly
established RO path. In this paper, we propose a scheme to solve the out-of-sequence problem with low cost.
In our scheme, we use the additional packet sequence number and the time information when the problem
occurs. We then run experiments on a reliable packet transmission (RPT) laboratory testbed to evaluate the
performance of the proposed scheme, and compare it with the well-known RO-supported PMIPv6 and the
out-of-sequence time period scheme. The experimental results show that for most of the cases, our proposed
scheme guarantees RPT by preventing the out-of-sequence problem.
INDEX TERMS PMIPv6, proxy mobile IPv6, route optimization.
I. INTRODUCTION
Internet usage continues its rapid expansion thanks to
the technological advances in wireless access technologies.
While overall IP traffic is expected to have 23 percent annual
growth between 2012 and 2017, IP traffic from mobile ter-
minals is expected to have 66 percent annual growth during
the same period [17]. IP mobility support has been a hot
topic over the last years, recently fostered by the role of
IP in the evolution of the 4G/LTE mobile communication
networks. Standardization bodies are working on different
aspects of the mobility aiming at improving the mobility
experience perceived by users. Having these requirements
of mobility in mind, IETF NETLMM WG has proposed
Proxy Mobile (PMIPv6) [1] as a new network-based mobility
protocol for IPv6 nodes which does not require host
involvements.
The main idea of PMIPv6 is that the Mobile Node (MN)
is not involved in any IP layer mobility-related signaling.
The MN is a conventional IP device (that is, it runs the
standard protocol stack). The purpose of PMIPv6 is to
provide mobility to IP devices without their involvement.
This provision is achieved by relocating relevant functions
for mobility management from the MN to the network.
This enables resource optimization in the networks and
reduces energy consumption of MN and handover signaling
cost. PMIPv6 performs better than MIPv6 in many aspects.
A Mobile Access Gateway (MAG) and a Local Mobility
Anchor (LMA) are in charge of the mobility of MN in
thePMIPv6 domain. However, basic PMIPv6 does not sup-
port Route Optimization (RO) and all messages that packets
related to MN are managed in LMA and MAG. In other
words, all packets are always transmitted via LMA, and
this increases the processing overhead of LMA as well as
transmission delay. Many schemes are proposed to support
the RO to overcome this problem [2]–[4].
When RO is supported in PMIPv6, the MN communicates
with the CN via the RO path between two different MAGs.
We define the RO path as a new path, and the basic PMIPv6
path as an old path. When the RO path is established, the out-
of-sequence problem occurs due to the different transmission
time between the old path and the RO path. This problem
causes packet loss in UDP and packet retransmission request
VOLUME 4, 2016
2169-3536
2016 IEEE. Translations and content mining are permitted for academic research only.
Personal use is also permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
1039
www.redpel.com +917620593389

B. Kang et al.: Developing RO-Based PMIPv6 Testbed for RPT
messages in TCP. These problems increase network overhead
and cannot provide reliable data transmission. We propose
a new scheme to solve that problem more accurately
and effectively. Our scheme uses the identical sequence
number of a packet and the original RO control message
of PMIPv6. Through the experimental measurement, pro-
posed scheme precisely prevents the out-of-sequence packets
compared to the OTP scheme, since it uses the sequence
number. In addition, it reduces the buffering cost by reducing
the number of entities. This paper introduce the extended
features and real test-bed measurement results of our previous
work [18].
The remainder of the paper is organized as follows.
Section 2 introduces the RO schemes in PMIPv6, the
EF-MIPv6 scheme, and the OTP scheme as related works.
In Section 3, we explain our proposed scheme. Section 4
presents performance modeling; we define the network
and mobility model and the equations for comparing the
amount of buffered packets and the packet reception delay.
In Section 5, we evaluate the number of out-of-sequence
packets and the handover delay via network simulation and
testbed measurement. Section 6 concludes the paper and
outlines our future work.
II. RELATED WORKS
PMIPv6 is starting to attract much attention among internet
communities and telecommunication due to its noble features
and it is expected to expedite the real deployment of
IP-based mobility management. However, an experimental
evaluation of PMIPv6, which analyzes the impact of its
practical constraints, is missing. In addition, the route
optimization problem is still challenging issue in handover
of mobile IP.
P. Loureiro proposed a PMIPv6 scheme [2] to support
the RO. When the MN attaches to the MAG domain,
MAG and LMA make a bidirectional tunnel by handshaking
the Proxy Binding Update (PBU) and Proxy Binding
Acknowledgement (PBA) message. After that, when the first
packet transmitted from the MN to the CN arrives at the LMA,
triggers the RO. The trigger message includes the MN-ID and
the MAG address. A new LMA, receiving the RO trigger
message, performs the RO control function. Normally, the
transmission delay in the RO path is less than the one in
the old path. In TCP network, the out-of-sequence prob-
lem causes frequent packet retransmissions. In case of UDP,
reliable service is not supported yet, although the packet
transmission delay is reduced by the RO in PMIPv6.
The Out-of-sequence Time Period (OTP) [5] scheme and
Enhanced Fast MIPv6 (EF-MIPv6) [6] are proposed to solve
the out-of-sequence problem. The OTP scheme is the solution
to restrain the tunnel establishment when the RO path is
established. The EF-MIPv6 scheme is one of the solutions
using the Enhanced Fast Binding Update (EF-BU) message
in MIPv6. However, the EF-MIPv6 scheme is not applicable
for PMIPv6, because EF-BU has to change the basic control
messages and works only in MIPv6. The OTP scheme cannot
provide reliable service to MN, because it is hard to predict
the restraint time of the tunnel in the OTP scheme.
Fast Handover Mobile IPv6 (FMIPv6) [7] was proposed to
reduce handover delay and packet losses in MIPv6. In their
algorithm, MN informs the new Access Router (nAR), as to
where MN would be, the previous Access Router (pAR) by
predicting its handover. Knowing the nAR’s address, the pAR
establishes a bidirectional tunnel with the nAR before the
MN attaches to the nAR. During MN’s handover, the pAR
forwards the packets to the nAR via the tunnel. When the nAR
receives the packets from the pAR, it stores all the packets
in the buffer. After the MN connects to the nAR, the nAR
forwards the buffered packets to the MN, and the packets
generated between the MN and the CN are forwarded via the
nAR without going through the pAR. If the distance between
the CN and the nAR is shorter than the tunneled distance from
the CN to the nAR via the pAR, the out-of-sequence problem
would occur. EF-MIPv6 was proposed to solve this problem.
The scheme in [6] uses the modified snoop protocol to
avoid the out-of-sequence problem in FMIPv6. In addition,
it solves the problem using Enhanced Fast Binding
Update (EF-BU) and the Multilink Procedure (MLP) mes-
sage and buffering at the nAR and the pAR. The procedure
for a handover in the scheme is similar to FMIPv6. How-
ever, the MN sends the EF-BU message to the CN after
the pAR exchanges the messages for setting up the tunnel
with the nAR to manage the out-of-sequence. Receiving the
EF-BU message, the CN changes the MLP field of the TCP
header in the message to 1 and then the message is forwarded
to the nAR. The nAR can determine if the packets come from
the pAR or the CN, by checking the number of the MLP field.
Then this resolves the out-of-sequence problem.
Even though the out-of-sequence problem in FMIPv6 is
solved in [6], it is hard to use this scheme in MIPv6. Because
the EF-BU and the MLP message have to be modified. More-
over, the buffering cost and the load of the routers increase
while buffering is performed at both the nAR and the pAR.
Finally, it is hard to apply the scheme in PMIPv6, because
PMIPv6 does not have the revised FMIPv6 protocol stack.
III. PROPOSED SCHEME
Our proposed scheme provides the reliable service for
MN more accurately to prevent the out-of-sequence prob-
lem while performing the RO. Our scheme manages the
problem effectively, using the packet sequence number, and
reducing the forwarding delay time using the value of
Time-To-Live (TTL).
A. MOTIVATION AND BASIC ASSUMPTIONS
The OTP scheme prevents the out-of-sequence problem
through the predetermined time of setting up the tunnel and
buffering between MAGs and LMA. In our proposed scheme,
only MAG performs buffering for MN so the buffering cost
in LMA decreases, whereas both the MAG and the LMA
in the OTP scheme. In addition, the OTP scheme does
not perfectly prevent the out-of-sequence problem due to
1040 VOLUME 4, 2016

the prediction. However, our proposed scheme solves the
out-of-sequence problem, using the packet sequence
numbers. In this paper, we assume that the MN sends packets
to CN after a MN’s handover between two domains to explain
this more effectively.
We use the IP header information to prevent the
out-of-sequence problem more effectively [8]. The identifica-
tion field, which is the number assigned from a router of the
IP header, is a unique number used by devising or recombin-
ing a packet following the Maximum Transfer Unit (MTU).
Accordingly, it is possible to know the packet sequence using
the identification number in the communication between
routers. MAGs and LMAs know the packet sequence via
the identification number in the IP header. Therefore, our
proposed scheme can determine the out-of-sequence packets
that arrive at the MAG by the identification number in the
IP header.
Our proposed scheme uses the TTL value in the IP header
to calculate the transmission time of the old path and the
RO path. We can count the number of routers through the
old path and the RO path from the TTL value. When a
packet passes through the tunnel, the tunnel header is added
to the packet. The TTL value in the tunnel header decreases
when the packet passes through the tunnel, since the packet
is encapsulated, but the TTL value in the IP header does
not. The packet is de-capsulated after passing through the
tunnel. Then, the TTL value in IP the header decrements
by one [9]. It is impossible to count the accurate number
of routers in each path due to this situation. Our scheme
uses the minimal encapsulation to avoid the problem [10]
by reducing the overhead of the tunnel header. The
TTL value usually decreases after a packet passes through the
tunnel, because the minimum information is kept at the inner
IP header, and the remaining information moves to the tunnel
header.
Our proposed scheme reduces the load of the router
by minimizing the number of routers that take buffering.
EntityNode is an entity connecting to a node. All of MAGs
and LMA perform buffering in the case of the OTP scheme;
however, our proposed scheme performs buffering on the
MAGCN. When a packet arrives at MAGCN through an old
path, MAGCN forwards the packet to CN. Conversely, if a
packet goes via the RO path, MAGCN stores the packet in
its buffer. Thus, our proposed scheme reduces the load of the
router and packet reception delay of the CN.
B. BASIC OPERATION
The procedure to establish the RO path is similar to
scheme [2]. The packets between MN and CN pass through
the old path before the RO path is established. If the
RO path is completely established, the packets pass through
the RO path. MAGCN receiving the packets via the RO path,
buffers the packets to cause the non-out-of-sequence prob-
lem. From the beginning of buffering in MAGCN, MAGCN
compares the sequence number of the first packet in the buffer
and the sequence number of the packet that passed via the
old path. MAGCN performs buffering until the last packet
from the old path arrives at MAGCN. When the last packet
from the old path arrives at MAGCN, MAGCN forwards the
packet and then forwards the all packets in its buffer.
In the proposed scheme, the out-of-sequence problem is
prevented by storing the packets from the RO path in the
MAG’s buffer, until all the packets pass from the old path.
Enabling the sequence number to understand the order of
all the packets passing though the old path and the RO path
resolves the problem more precisely than other schemes do.
In addition, our proposed scheme transfers via the shortest
path due to performing buffering in MAGCN. Thus the out-of-
sequence problem is not occurred, and the packet reception
delay is reduced after the RO path is established. MAGCN
forwards the packets in the buffer to CN after the last packet
from the old path passes through MAGCN. However, if the
last packet from the old path is lost, MAGCN performs buffer-
ing infinitely. The maximum forwarding delay time (Twait)
is calculated to prevent infinite buffering in our proposed
scheme. If the last packet from the old path does not arrive
at MAGCN within the Twait, the packets in the buffer are
forwarded to CN. The out-of-sequence problem is prevented
using the maximum forwarding delay time, even though the
packets from the old path are lost.
Twait is used to prevent infinite buffering. Twait is calcu-
lated by the time difference between the times that the packets
coming from the old path and the RO path, arrive at MAGCN.
Twait is calculated by equation (1). TOP defines the time
that the packet passes via the old path; and TNP defines the
time the packet passes via the RO path. From equation (1),
we can calculate the different arrival times between
TOP and TNP.
Twait = TOP − TNP (1)
Equation (2) and (3) are the formulas to calculate
TOP and TNP. TTLMax is the maximum value of TTL.
TTLOP and TTLNP are the TTL values of the packets from
the old path and the RO path, respectively. TTLMax minus
TTLOP or TTLNP calculates the number of routers that a
packet has passed in each path. We can calculate the packet
transmission time of each path by multiplying the counted
number of the routers with the packet transmission time per
hop, TOne−Hop.
TOP = (TTLMax − TTLOP) · TOne−Hop (2)
TNP = (TTLMax − TTLNP) · TOne−Hop (3)
TOne−Hop is calculated by equation (4). TOne−Hop is
calculated using the round-trip time and the TTL value of the
RO Setup message that sets up the RO path. The Ethernet’s
MTU is used to calculate the maximum forwarding delay
time. TTLRS is the TTL value when the RO Setup message
arrives at MAGCN, and LMTU is the MTU size the network
has. TRS is the round-trip time of the RO Setup message and
LRO-Setup is the size of the RO Setup message. The transmis-
sion time of the old path, the RO path, and the transmission
VOLUME 4, 2016 1041

time per hop are calculated using the RO Setup message.
TOne−Hop =
LMTU ·
TRS
2
LRO−Setup
TTLMax − TTLRS
(4)
C. SIGNALING OF THE PROPOSED SCHEME
In our proposed scheme, the signaling flow for RO is similar
to scheme [2]. The signaling to solve the out-of-sequence
problem is performed in MAGCN. Figure 1 shows the sig-
naling flow of our proposed scheme when the last packet of
the old path is lost in MN’s inter-domain handover. When
MAG2 receives the RO Init message from LMA2, MAG2
performs the flow to calculate t Twait. MAG2 saves the TTL
value of the RO Setup message and inter-arrival time of the
RO Setup message between MAG1 and MAG2. When the
RO is completed, MAG2 starts to store the packets from
the RO path and checks the packet sequence number from
the old path. If the last packet from the old path arrives at
MAG2, MAG2 forwards the packet to CN. Next, MAG2
forwards all the buffered packets in MAG2 to CN. However,
if the last packet from the old path does not arrive at MAG2
during Twait, MAG2 thinks the last packet from the old path
is lost. Therefore, MAG2 forwards all buffered packets in
MAG2 to CN.
FIGURE 1. Signaling flow of proposed scheme.
Figure 2 is the flow chart of the algorithm in MAG2.
The flow chart is performed after the RO path is established
between MAG1 and MAG2. Then, MAG2 receives the first
packet from the RO path. MAG2 starts to check the elapsed
time by the time of starting buffering. When the packet arrives
at MAG2, MAG2 checks the path using the packet. If the
packet arrives from the RO path, MAG2 stores the packet
in its own buffer. If the packet arrives from the old path,
MAG2 compares the packet sequence with the packet and the
first packet in the buffer. If the last packet from the old path
arrives at MAGCN, the packets in the buffer are forwarded
to CN after MAGCN forwards the last packet from the old
path to CN. However, if the last packet from the old path
does not arrive at MAG2 during Twait, MAG2 forwards all
buffered packets in MAG2 to CN. Our proposed scheme
provides reliable service to avoid the out-of-sequence more
FIGURE 2. Flow chart of the proposed scheme.
precisely than the OTP scheme. The out-of-sequence problem
is prevented using Twait, even though the packets from the old
path are lost. In addition, the scheme minimizes the end to end
packet reception delay, using the old path during the estab-
lishment of the RO path. Moreover, our proposed scheme
reduces the buffering cost, because buffering is performed
only by MAGCN.
IV. PERFORMANCE MODELING
In this section, we use mathematical modeling to compare
the performance of our proposed scheme to the OTP scheme.
In section 4.1, we define the network model to use in the per-
formance modeling. In section 4.2, we define the equations
to compare the amount of buffered packets and the packet
reception delay in our proposed scheme and the OTP scheme.
Section 4.3 shows the results of the performance evaluation.
We developed initial version of Proxy Mobile IPv6 and its
analytical model in 2012 [18].
A. NETWORK AND MOBILITY MODEL
Figure 3 is the network topology for our performance model-
ing. The network model has two LMAs; each LMA connects
FIGURE 3. Network topology for the performance modeling.
1042 VOLUME 4, 2016

with two MAGs. dX-Y denotes the number of hops between
entity X and entity Y. We use the fluid-flow mobility model,
which is popular to use in performance evaluation of mobile
network, shows mobility of MN considering the velocity and
direction of MN [11], [12]. Our proposed scheme calculates
the rate of the MN’s handover using the fluid-flow model.
The fluid-flow model calculates the average number of MNs
that perform the handover by utilizing the number of MAGs
composed in the LMA domain and the session arrival rate
of MN.
µc, the rate of MN’s intra-domain handover, is calculated
by equation (5). ν is the average velocity of MN, R is the
radius of MAG’s coverage, and S is the area of MAG’s
coverage. µd , the rate of MN’s inter-domain, is calculated
by equation (6). N is the number of MAGs in LMA domain.
µs, the rate of the intra-domain handover, is calculated by
equation (7) [11]–[13].
µc =
2 · v
√
π · S
=
2 · v
π · R
(5)
µd =
µc
√
N
(6)
µs = µc − µd = µc ·
√
N − 1
√
N
(7)
The average number of MNs is determined using λs in
fluid-flow model mobility; that is, session arrival rate, µc,
µd , and µs. E[Nc], the average number of MNs that do the
handover is calculated by equation (8). E[Nd ], the average
number of MNs in the inter-domain handover, is calculated
by equation (9), and E[Ns], the average number of MNs in
the intra-domain handover, is calculated by equation (10).
E [Nc] =
µc
λs
(8)
E [Nd ] =
µd
λs
(9)
E [Ns] =
µs
λs
=
(µc − µd )
λs
(10)
B. MATHEMATICAL MODELING
The packet reception delay denotes that a CN receives the first
packet from MN after handover. If the packet reception delay
is short, CN realizes that the handover delay is short. The
OTP scheme restricts the disestablishment of the old path
until the RO path is established. Therefore, CN receives the
first packet from MN after finishing the binding update for
MN and establishing the RO path. In our proposed scheme,
CN receives the first packet from MN after finishing the
binding update for MN. The packet reception delay of our
proposed scheme is shorter than for the OTP scheme due to
this feature.
DOTP, the packet reception delay in the OTP scheme, is
calculated by equation (11). DPBU and DPBA are the trans-
mission delay of the PBU and PBA messages, respectively.
DRO-Inter and DRO-Intra are the delay of the RO setup in
the inter-domain and intra-domain, respectively. DNP-Data is
the packet transmission delay from MAGMN to CN after the
RO [14], [15].
DOTP = (DPBU + DPBA) +
µd
µc
· DRO−Inter + 1 −
µd
µc
· DRO−Intra] + DNP−Data (11)
DPBU and DPBA are calculated by equations (12) and (13),
respectively. α is the transmission cost in a wired network,
and Li is the size of message i. Pt is the packet processing
delay in a router. PLMA and PMAG are the packet processing
delay in LMA and MAG, respectively. Bw is the bandwidth
in a wired network.
DPBU = α ·
LPBU
Bw
+ Pt · (dMAG−LMA − 1) + PLMA
(12)
DPBA = α ·
LPBA
Bw
+ Pt · (dMAG−LMA − 1) + PMAG
(13)
DRO-Inter and DRO-Intra are calculated by equations (14).
Di is the transmission delay of message i. The transmission
delay of each message is calculated by equations (15)-(20).
DRO-Intra = DRO-Init + DRO−InitAck
+ DRO−Setup + DRO−SetupAck (14)
DRO−Trigger = α ·
LRO−Trigger
Bw
+ Pt
· (dLMA−LMA − 1) + PLMA (15)
DRO−TriggerAck = α ·
LRO−TriggerAck
Bw
+ Pt
· (dLMA−LMA − 1) + PLMA (16)
DRO−Init = α ·
LRO−Init
Bw
+ Pt
· (dMAG−LMA − 1) + PMAG (17)
DRO−InitAck = α ·
LRO−InitAck
Bw
+ Pt
· (dMAG−LMA − 1) + PLMA (18)
DRO−Setup = α ·
LRO−Setup
Bw
+ Pt
· (dMAG−MAG − 1) + PMAG (19)
DRO−SetupAck = α ·
LRO−SetupAck
Bw
+ Pt
· (dMAG−MAG − 1) + PMAG (20)
DNP−Data is the packet reception delay from MN to a CN
via the RO path. DNP−Data is calculated by equation (21).
Sd and τ are the size of a packet and a tunnel header, respec-
tively. β is the transmission delay in the wireless network, and
Bwl is the network bandwidth.
DNP−Data = α ·
sd + τ
Bw
+ Pt · dMAG−MAG
+ β ·
sd
Bwl
+ Pt · dMN−MAG (21)
VOLUME 4, 2016 1043

DPro, the packet reception delay in our proposed scheme,
is calculated by equation (22). DPro is calculated by adding
the delay of the binding update, the data transmission delay
from MN to a CN, and the extra delay to lose the last
packet from the old path. DOP_Inter-Data and DOP_Intra-Data
are the data transmission delay through the old path in
the inter-domain handover and the intra-domain handover,
respectively.
DPro = (DPBU + DPBA)
+
µd
µc
· DOP_Inter−Data + 1−
µd
µc
·DOP_Intra−Data
+ λd ·
µd
µc
· TWait−Inter + 1 −
µd
µc
· TWait−Intra
(22)
Twait−Inter and Twait−Intra, the maximum forwarding delay
of the RO path of the intra-domain handover and inter-
domain handover, respectively, are calculated by the chang-
ing value of TTLRO−Setup in formula Twait. We consider the
worst delay to lose the last packet from the old path, to
measure the accurate packet reception delay in the proposed
scheme. DOPI nter−Data and DOPI ntra−Data are calculated by
equations (23) and (24), respectively.
DOP_Inter−Data = 2β ·
sd
Bwl
+ Pt · dMN−MAG
+ 2α ·
sd + τ
Bw
+ Pt · dMAG−LMA
+ α ·
sd
Bw
+ Pt · dLMA−LMA (23)
DOP_Intra−Data = 2β ·
sd
Bwl
+ Pt · dMN−MAG
+ 2α ·
sd + τ
Bw
+ Pt · dMAG−LMA
(24)
The amount of buffered packets is the total number of
stored packets in the buffer at LMA and MAG to pre-
vent packet loss after MN’s handover. Our proposed scheme
reduces the amount of buffered packets, because the packets
are only buffered at MAG, since the OTP scheme stores the
packets in both the MAG and t LMA.
The amount of buffered packets in the OTP scheme is cal-
culated by equation (25). SOTP-Buffer is the amount of buffered
packets when the RO path is set up in the OTP scheme.
λp, the generation rate of data traffic, expresses the number
of the transmitted packets per unit time [13]. The amount
of buffered packets in our proposed scheme is calculated
by equation (26). Our proposed scheme sends the packets
through the old path during the RO path set up. Therefore,
the time to start buffering is the time when the RO path
establishment is completed. The maximum time to buffer the
packets is Twait-Inter .
SOTP−Buffer = E (Nd ) · DRO−Inter · λp
+ E (Ns) · DRO−Intra · λp (25)
SPro−Buffer = E (Nd ) · TWait−Inter · λp
+ E (Ns) · TWait−Intra · λp (26)
C. PERFORMANCE EVALUATION
We define the equations in mathematical modeling to com-
pare the amount of buffered packets and the maximum packet
reception delay of our proposed scheme and the OTP scheme.
In this section, we compare the performance of our proposed
scheme to the OTP scheme using the parameter value in
Table 1 [13]–[15]. Figure 4 shows the packet reception delay
impacted by the changes of the data size. N is 25, and
ν is 20 m/s. λs and λd is 0.1. Sd increases from 100 bytes
to 1,500 bytes and MTU is set to 1,500, as the general
value of the Ethernet. Mathematical modeling shows the
OTP scheme incurs a longer packet reception delay than
our proposed scheme does. As the data size increases, the
difference between the two schemes lessens, but our proposed
scheme has better performance, because it is hard to exceed
a 1,500 byte transfer in general Ethernet.
Figure 5 shows the packet reception delay impacted by the
changes in the number of MAGs in LMA domain. We set
ν, λs, λd , and MTU as the same value, as in the above
environment.
Sd is set to 500 and 1,500 bytes, and N increases
from 1 to 50. Increasing the number of MAGs in LMA
domain decreases the rate of inter-domain handover. There-
fore, the packet reception delay decreases, as the number of
MAGs increase. The reception delay in our proposed scheme
is shorter than that of the OTP scheme.
Figure 6 shows the packet reception delay impacted by
losing the last packet from the old path. We set N, ν, λs, and
MTU is the same value as in the environment of the result in
figure 6. We set Sd to 500 bytes and λd increases from 0 to 1.
As the probability that the loss of the last packet from the old
path increases, the probability also increases that the extra
delay is as much as Twait. Thus, the packet reception delay
increases as the probability that the last packet is lost in the
old path. However, the packet reception delay in our proposed
scheme is shorter, even though the last packet from the old
path is always lost.
Figure 7 shows the amount of buffered packets impacted by
the sending traffic rate in MN. N is set to 25, ν is set to 20 m/s,
and λs is set to 0.1. λp increases from 1 Mbps to 15 Mbps.
As the sending traffic rate increases, the OTP scheme stores
more packets in the buffer than our proposed scheme does
and the difference increases.
Figure 8 shows the amount of buffered packets impacted by
the average velocity of MN. N is set to 25, λs is set to 0.1, and
λp is set to 3 Mbps. ν increases from 1 m/s to 50 m/s. As MN’s
average velocity increases, the number of t MNs that perform
handover increase. Thus, the amount of buffered packets also
1044 VOLUME 4, 2016

TABLE 1. Parameter values for the performance evaluation.
FIGURE 4. Packet reception delay impacted by the changes the data size.
increases, since the number of the RO path establishments
increases. If the average velocity of MN increases, the OTP
scheme stores the packet in the buffer more than our proposed
scheme does.
FIGURE 5. Packet reception delay impacted by the changes the number of
MAGs in LMA domain.
Figure 9 shows the amount of buffered packets changes
with the number of MAGs in LMA domain. λs is set to 0.1,
λp is set to 3 Mbps, and ν is set to 20 m/s. N increases from
1 to 35. As the number of MAGs in LMA domain increases,
VOLUME 4, 2016 1045

FIGURE 6. Packet reception delay impacted by losing the last packet in
the old path.
FIGURE 7. Amount of the buffered packets impacted by the sending
traffic rate.
FIGURE 8. Amount of the buffered packets impacted by the average
velocity of MN.
the number of the inter-domain handovers increases. Thus,
the amount of buffered packets decreases, as the number
of MAGs increases. The amount of buffered packets in our
proposed scheme is smaller than for the OTP scheme, and
buffering is performed effectively, as the number of MAGs
decreases.
V. SIMULATION RESULTS
In this section, we verify the number of out-of-sequence
packets and the packet reception delay through the com-
puter simulation and testbed experiment. We check the
number of out-of-sequence packets in the RO supported
PMIPv6, OTP scheme, and our proposed scheme using the
FIGURE 9. Amount of the buffered packets impacted by the number of
MAGs in LMA domain.
comprehensive computer simulation. We also implement a
testbed for the RO supported PMIPv6 and verify algorithms
in real environment. Our proposed scheme generates fewer
out-of-sequences than the OTP scheme and the RO supported
PMIPv6 does. In addition, it demonstrates improved perfor-
mance in terms of packet transmission delay compared to the
OTP scheme. We illustrate the simulation results in the former
part (5.1 and 5.2) and the testbed measurements are illustrated
in later part (5.3 and 5.4).
FIGURE 10. Network topology of simulation.
A. SIMULATION ENVIRONMENT
We run the simulator implemented in C++ to measure
the number of the out-of-sequence packets. We conduct
our simulation in the UDP environment to determine the
packet loss and out-of-sequence packets. The simulation
uses the CBR traffic generator, and data packets are gener-
ated in 0.02 seconds with the size of 500bytes. Using the
CBR traffic, we can verify the incidence of packet trans-
mission delay and the number of out-of-sequence packets
accurately. Figure 10 is the network topology to conduct the
simulation. We configure dLMA−LMA as 15 hops, dMAG−LMA
as 7 hops, dMAG−MAG as 7 hops, dMN−MAG as 1hop. In this
simulation, the number of the out-of-sequence packets and
the packet reception delay are verified during establishment
of the RO path in the inter-domain handover.
B. SIMULATION RESULT
Figure 11 shows the simulation results of the RO supported
PMIPv6. This scheme cannot prevent occurrence of
out-of-sequence packets; 66 out-of-sequence packets occurred.
1046 VOLUME 4, 2016

FIGURE 11. RO supported PMIPv6.
FIGURE 12. Simulation results while changing OTP value. (a) OTP = 0.5.
(b) OTP = 0.8.
In addition, 49 packets were lost. It incurred 0.0947 seconds
of packet reception delay.
Figure 12 shows the simulation results when OTP is set
to 0.5 (a) and 0.8 (b), respectively, in the OTP scheme.
41 out-of-sequence packets occurred when OTP was set to
0.5. 50 packets were lost and it incurred 1.2 seconds of packet
reception delay during handover. When OPT was set to 0.8,
30 out-of-sequence packets occurred. 50 packets were lost
and it incurred 1.82 seconds of packet reception delay dur-
ing handover. The out-of-sequence packets decreased when
OTP increases, but the packet reception delay increased.
Figure 13 is the simulation results of our proposed scheme.
The results show the cases where the last packet from the
old path are lost and not lost, respectively. Out-of-sequence
FIGURE 13. Simulation results of proposed scheme. (a) Not loss. (b) Loss.
TABLE 2. Simulation results.
will not be incurred, if no packets are lost from the old path.
The packets lost and reception delay are the same as in the
RO supported PMIPv6. Even if the last packet from the old
path are lost, the out-of-sequence packets will not occur,
but approximately 1.91 seconds of receiving packet delay is
incurred, because it transmits stored packets in the buffer, as
much as Twait.
Table 2 shows the simulation results of each scheme. The
three compared schemes in the simulation incurred a similar
number of lost packets, because they do not have a func-
tion to prevent packet loss. The OTP scheme decreases the
packet loss compared to the RO supported PMIPv6, but it
increases reception delay. However, it prevents all out-of-
sequence packets when the RO path is established, because
our proposed scheme uses the packet sequence number. If the
last packet from the old path is not lost, the delay time is the
same as in the RO supported PMIPv6. Our proposed scheme
VOLUME 4, 2016 1047

prevents the out-of-sequence problem and supports reliable
service more effectively.
C. TESTBED ENVIRONMENT
We establish the PMIPv6 testbed and do experimental work to
observe the performance. Our testbed is based on Open Air
Interface (OAI) PMIPv6 v0.3 [16]. To establish the testbed
for the RO supported PMIPv6 and our proposed scheme, we
use Ubuntu 10.04 and C language. Figure 14 represents the
topology of our testbed. It includes 2 LMAs, 3MAGs, and
MN and CN. MAG1 and MAG2 are connected to LMA1, and
MAG3 is connected to LMA2. The MN and CN are connected
to MAG1 and MAG3 respectively. The MN sends 200 packets
to the CN in a second. Then the MN roams to MAG2 domain,
which causes handover latency of 1 second. We use a hub
rather than an AP for the wireless link, and connect the MN to
MAG with a cable. Although we can obtain the same results in
a wireless environment as in a wired environment, the reason
to use a cable is to minimize the signal interference and the
effect of signaling size in a wireless environment.
FIGURE 14. Network topology of testbed.
We also regard unified handover time by clarifying the
start and end points of the handover in each case within the
same environmental conditions. We develop the RO module
to establish the RO supported PMIPv6 testbed. In the mod-
ule, the messages for route optimization such as RO trigger,
RO init, and RO Ack, are defined. The LMAs also transfer
another message, which is also newly defined for LMAs to
share information of MNs in their domain, with each other
when MN attaches. We need to use this message because the
LMAs in the RO scheme are assumed to share information
of MNs in their domain each other. We add the tunneling and
packet routing functions for RO path between MAGs to the
module.
To establish the testbed for our proposed scheme, MAGs
needs packet buffering function. For this, we implement the
packet buffering module which uses Netfilter and IP6Tables
to hook packets. The Netfilter is a packet filtering framework
embedded in Linux kernel 2.4.x and 2.6.x version. It pro-
vides hook handlings or hook points for intercepting and
manipulating network packets. IP6Tables utility is a tool in
FIGURE 15. RO supported PMIPv6.
FIGURE 16. Testbed results of proposed scheme. (a) Not loss. (b) Loss.
user space to provide hook handlers for the hook points of
Netfilter. The packet buffering module adds a rule to the
IP6Tables to decide which packet should be buffered, and
hooks packets through the RO path by using the Netfilter.
It also stores the hooked packets by using Libipq library in
IP-Tables tool in user space. We finish the implementation of
our proposed scheme by adding packet forwarding function,
which performs packet forwarding when the MAG receives
the last packet through the old path or buffering time, which
have been set in advance is over. In our testbed, the MAG3
connected to the CN performs buffering packets from the
RO path.
D. TESTBED RESULT
Figure 15 shows the testbed results of the RO supported
PMIPv6. This scheme cannot prevent out-of-sequence pack-
ets. We disconnect the MN from the MAG1 and move it to
1048 VOLUME 4, 2016

the MAG2. The handover latency is 1 second and 178 packets
are lost during the MN’s handover. After the MN’s handover,
the MN begins to receive packets, but 78 out-of-sequence
packets are generated right after the RO path is established.
As referring to Figure 15, we can see that the RO supported
PMIPv6 generates out-of-sequence packets due to the gap of
packet transmission latency between the old and RO path in
real environment.
Figure 16 is the testbed results of our proposed scheme.
The simulation results show the cases where the last packet
from the old path is lost and where not lost. Through
Figure 16(a), we can see that there are no out-of-sequence
packets unless packets are lost in the old path. The imple-
mented packet buffering module performs buffering packets
through the RO path, and forwards them when the last packet
through the old path arrives. When packets are lost from
the old path, the reception delay for 1.88 seconds occurs as
Figure 16(b) represents. That is because the MAG should wait
for Twait to decide whether packets are lost or not from the old
path. However, the out-of-sequence packets are not generated
during the measurement time.
VI. CONCLUSION
The difference of the transmission delay between the old path
and the RO path generates the out-of-sequence problem in
PMIPv6. To solve this problem, several types of literature
suggested sophisticate algorithms, but they did not solve the
problem entirely. For that reason, we proposed a new algo-
rithm that provides reliable service for MN more accurately.
We use the packet sequence number that reduces the forward-
ing delay time. To evaluate our scheme, we compare with the
well-known RO supported PMIPv6 and the OTP scheme via
computer simulation and testbed measurement. Our proposed
scheme solved the out-of-sequence problem in both case
(simulation and system measurement). Furthermore, we saw
that our scheme reduced the packet reception delay after the
RO path is established. It can be possible to provide a reliable
service in PMIPv6 RO by adapting our proposed scheme.
REFERENCES
[1] S. Gundavelli, K. Leung, V. Devarapalli, K. Chowdhury, and B. Patil, Proxy
Mobile IPv6, Standard IETF RFC 5213, Aug. 2008.
[2] P. Loureiro and M. Liebsch, Proxy Mobile IPv6 Localized
Routing, documents IETF RFC, Jan. 2010. [Online]. Available:
https://tools.ietf.org/html/draft-loureiro-netext-pmipv6-ro-00
[3] S. Krishnan, R. Koodli, P. Loureiro, Q. Wu, and A. Dutta. (Sep. 2012).
Localized Routing for Proxy Mobile IPv6. [Online]. Available:
http://www.hjp.at/doc/rfc/rfc6705.html
[4] B. O. Obele, J. K. Rhee, and M. Kang, ‘‘A proposal for reducing handover
latency and improving route optimization in proxy mobile IPv6,’’ in Proc.
10th ICACT, Feb. 2008, pp. 49–54.
[5] J.-H. Lee, Y.-D. Kim, and D. Lee, ‘‘Tunnel restraint to prevent out-of-
order packets for route optimization in proxy mobile IPv6,’’ Wireless Pers.
Commun., vol. 60, no. 3, pp. 547–558, Apr. 2011.
[6] B. Park and H. Latchman, ‘‘Performance enhancement of fast handover for
MIPv6 by reducing out-of-sequence packets,’’ Wireless Pers. Commun.,
vol. 47, no. 2, pp. 207–217, Mar. 2008.
[7] R. Koodli, Fast Handovers for Mobile IPv6, Standard RFC 4068, Jul. 2005.
[8] J. Postel, Internet Protocol, Standard RFC 791, Sep. 1981.
[9] C. Perkins, IP Encapsulation Within IP, Standard RFC 2003, Oct. 1996.
[10] C. Perkins, Minimal Encapsulation within IP, Standard RFC 2004,
Oct. 1996.
[11] X. Zhang, J. Castellanos, and A. Campbell, ‘‘P-MIP: Paging extensions for
mobile IP,’’ Mobile Netw. Appl., vol. 7, no. 2, pp. 127–141, Apr. 2002.
[12] F. V. Baumann and I. G. Niemegeers, ‘‘An evaluation of location man-
agement procedures,’’ in Proc. 3rd Annu. IEEE Int. Conf. Univ. Pers.
Commun., Sep./Oct. 1994, pp. 359–364.
[13] C. Makaya and S. Pierre, ‘‘An analytical framework for performance
evaluation of IPv6-based mobility management protocols,’’ IEEE Trans.
Wireless Commun., vol. 7, no. 3, pp. 972–983, Mar. 2008.
[14] J. H. Lee, T. Ernst, and T. M. Chung, ‘‘Cost analysis of IP mobility man-
agement protocols for consumer mobile devices,’’ IEEE Trans. Consum.
Electron., vol. 56, no. 2, pp. 1010–1017, May 2010.
[15] S. Lo, G. Lee, W.-T. Chen, and J.-C. Liu, ‘‘Architecture for mobility and
QoS support in all-IP wireless networks,’’ IEEE J. Sel. Areas Commun.,
vol. 22, no. 4, pp. 691–705, May 2004.
[16] EURECOM. EURECOM Open Air Interface Proxy Mobile IPv6, accessed
on Mar. 2016. [Online]. Available: http://www.openairinterface.org/
[17] Cisco Company, ‘‘Cisco visual networking index: Forecast and methodol-
ogy,’’ Cisco, San Jose, CA, USA, White Paper C11-481360, May 2015.
[18] N. Kwon, M. Kim, S. Oh, and H. Choo, ‘‘The reliable packet transmission
based on PMIPv6 route optimization,’’ in Proc. IEEE 75th Veh. Technol.
Conf. (VTC Spring), May 2012, pp. 1–4.
BYUNGSEOK KANG received the B.S. degree
in computer engineering from Sejong University,
Korea, in 2006, the M.S. degree in electrical and
electronics engineering from Korea University,
Korea, in 2008, and the Ph.D. degree in elec-
tronics and computer science from the University
of Southampton, U.K., in 2015. He is currently
a Research Professor with Sungkyunkwan Uni-
versity in 2015. His research interests include
wired/wireless networking, sensor networking,
mobile computing, network security protocols, and simulations/numerical
analysis.
NAMYEONG KWON received the B.S. degree
in computer engineering from Kangwon National
University, and the M.S. degree from the College
of Information and Communication Engineering,
Sungkyunkwan University, Korea, in 2012. He is
currently doing research with Samsung Electron-
ics, Korea. His research interests include proxy
mobile IPv6, route optimization, mobile comput-
ing, and wireless communication.
HYUNSEUNG CHOO (M’99) received the
B.S. degree in mathematics from Sungkyunkwan
University, Korea, in 1988, the M.S. degree in
computer science from the University of Texas
at Dallas, USA, in 1990, and the Ph.D. degree in
computer science from the University of Texas at
Arlington, USA, in 1996. From 1997 to 1998, he
was a Patent Examiner with the Korean Industrial
Property Ofﬁce. Since 1998, he has been with the
College of Information and Communication Engi-
neering, Sungkyunkwan University, and has been an Associate Professor and
the Director of the Convergence Research Institute. Since 2005, he has been
the Director of the Intelligent HCI Convergence Research Center (eight-
year research program) supported by the Ministry of Knowledge Economy
(Korea) under the Information Technology Research Center support program
supervised by the Institute of Information Technology Assessment. He has
authored over 200 papers in international journals and refereed conferences.
His research interests include wired/wireless/optical embedded networking,
mobile computing, and grid computing. He is member of ACM. He is
the Vice President of the Korean Society for Internet Information (KSII).
He has been the Editor-in-Chief of the Journal of KSII for three years
and journal editors of the Journal of Supercomputing, the International
Journal of Mobile Communication, and Springer-Verlag Transactions on
Computational Science Journal, and an Editor of KSII Transactions on
Internet and Information Systems since 2006.
VOLUME 4, 2016 1049

Developing route optimization based pmi pv6 testbed for reliable packet transmission

More Related Content

What's hot (19)

Viewers also liked (14)

Similar to Developing route optimization based pmi pv6 testbed for reliable packet transmission (20)

More from redpel dot com (20)

Recently uploaded (20)

Developing route optimization based pmi pv6 testbed for reliable packet transmission