ADVANCE DUPLICATE ADDRESS DETECTION (DAD) AND
BUFFERING-BASED PROVISION OF SEAMLESS HANDOVER
IN MOBILE IPV6
Shobhan Adhikari, Teerapat Sanguankotchakorn
Telecommunications Field of Study, School of Advanced Technologies, Asian Institute of Technology, P.O. Box 4, Klong
Luang, Pathumthani 12120, Thailand
Keywords: Seamless Handover, Mobile IPv6, Buffering, Simulation, NS-2, Smooth Handover, Fast Handover,
Performance Analysis.
Abstract: Due to the concerns about the imminent crunch in available addresses in the previous version of Internet
Protocol, IPv4, and to offer additional functionalities for new devices, IPv6 was proposed. Mobile IPv6
(MIPv6), an extension to IPv6, manages Mobile Nodes’ movements between wireless IPv6 networks. One
of the most important considerations for Mobile IPv6 is handover management. It is desired that handover
be fast and lossless. Seamless handovers are such that they incur minimum packet loss and delay. Various
proposals have been made for seamless handover in MIPv6. By forwarding the packets destined to the
Mobile Node towards the new point of attachment and storing the packets there until the Mobile Node has
attached there, packet loss can be significantly decreased, and the delay associated with the forwarding is
also less compared to forwarding from the previous point. In this paper, we study the performance of one
such scheme which has optimized fast handover over hierarchical structure with buffering and simulation
using NS-2 to evaluate packet loss and delay for UDP streams. It was observed that with the buffering
scheme used, the handover was seamless. There was a difference in latencies with and without handover, as
expected. It was observed that most of the performance factors studied depended on the data rate of the
traffic. The factors were found to be more dependent on the data rate than on the speed of the Mobile Node.
1 INTRODUCTION
With the increasing use of Internet capable and
handheld wireless devices, the requirement of
seamless handover has driven various proposals
towards decreasing the delay and loss associated
with handover.
In Mobile IPv6, each time Mobile Node (MN)
moves from one subnet to another, it gets a new
Care-of Address (CoA). After obtaining a CoA, it
registers Binding, consisting of its new Care-of
Address, the home address and the registration
lifetime, with the Home Agent (HA) and the
Correspondent Node(s) (CN(s)) it is communicating
with. In case of CN without MN binding, packets
reach HA and from there are tunneled to CoA,
whereas CN with knowledge of the Binding can
send the packets directly to MN’s CoA. As the
number of MNs increases and cell sizes start to
shrink to increase the capacity, number of Binding
Updates increases proportionately, causing a
significant signaling overhead. For solving this
problem, Hierarchical Mobile IPv6 (HMIPv6) was
proposed (Castelluccia, 1998). In HMIPv6, Mobility
Anchor Point (MAP), router highest in the hierarchy
in the visited network, acts like a local HA for the
visiting MN. It also limits the amount of signaling
required outside MAP's domain. The hierarchical
scheme separates local mobility (micro-mobility)
from regional mobility (macro-mobility). MN
changes only the Local Care-of Address (LCoA)
inside a local domain and not the Regional Care-of
Address (RCoA). Packets addressed to MN’s RCoA
are routed to the subnet, intercepted by MAP, and
tunneled to MN’s LCoA. This scheme improves
handover performance and reduces signaling load.
With fast handover (Koodli, 2003), the delay
involved with handover is reduced, the latency being
comparable to L2 handover latency. It reduces
packet loss by providing fast IP connectivity as soon
as MN changes to a new point of attachment. Fast
Handover may either be Tunnel-based or
Anticipated Handovers. In case of Tunnel-based
44
Adhikari S. (2005).
ADVANCE DUPLICATE ADDRESS DETECTION (DAD) AND BUFFERING-BASED PROVISION OF SEAMLESS HANDOVER IN MOBILE IPV6.
In Proceedings of the Second International Conference on e-Business and Telecommunication Networks, pages 44-52
DOI: 10.5220/0001418500440052
Copyright
c
SciTePress
Handover, routing is fixed during link configuration
and binding update, so that packets delivered to the
Old CoA (OCoA) are forwarded to the New CoA
(NCoA) by setting up a bidirectional tunnel between
old access point and new access point. In
Anticipated Handover, FMIPv6 provides support for
pre-configuration of link information (such as the
subnet prefix) in the new subnet while MN is still
attached to the old subnet, by the use of L2 triggers.
This reduces the amount of pre-configuration time in
the new subnet. Fast Handover scheme is analyzed
in (Pack and Choi, 2003). For better results,
Hierarchical structure and Fast Handover can be
used together (Jung et al., 2004). According to
(Perez-Costa et al., 2003), performance of
combination of HMIPv6 and FMIPv6 is better than
either of them acting alone.
But even with fast handover, there still is a
probability of packet loss. Bi-casting is a viable
scheme wherein MAP performs bi-casting to both
Previous Access Router (PAR) and New Access
Router (NAR), so that they both initiate sending of
packets to MN’s OCoA and NCoA in response to
handover indication. But bi-casting is not very
efficient because of the overhead involved with
sending packets to both addresses, causing
redundancy. To decrease redundancy, coordinated
bi-casting could be performed. In such a scheme,
NAR and PAR agree on a switching point, which
defines the exact moment for switching service from
PAR to NAR.
The various factors contributing to handover
delays are: IP address assignment when DHCP
server is far from MN (in case of Stateful (Bound et
al., 2001) Auto-Configuration), Duplicate Address
Detection (DAD) and Neighbor Discovery (ND) are
the main contributors to the latency. The other
contributors are the various signaling. Handover
latency can be decreased if the delays due to the
factors mentioned can be decreased. After forming a
new CoA, with either Stateless (Thomas and Narten,
1998) or Stateful Auto-configuration, MN may
perform DAD on it. For DAD, MN sends one or
more Neighbor Solicitations to its new address and
waits for a response for at least one second, hence
contributing a significant portion to the total
handover delay (Montavont and Noel, 2003). Hence,
with some scheme to reduce this delay, the overall
handover delay could be reduced significantly. (Lee
et al., 2001) has a scheme where MN performs DAD
while using OCoA and packets are buffered at PAR
during handover. With this, packets have to travel
the distance PAR-MAP-NAR while they are
forwarded to MN. To improve performance further,
buffering could be done at NAR so that buffered
packets can avoid the additional distance of PAR-
MAP.
Advance DAD (Han et al., 2003) can be used to
reduce the delay contributed by DAD to the overall
handover delay. Advance DAD scheme
automatically allocates a globally routable IPv6 CoA
for the use of MNs that participate in fast handover.
Buffering reduces packet loss by storing packets
destined for MN during the time MN is handing over
and forwarding the same to MN after it has
established link connectivity.
2 RELATED WORKS
In this section, some related works for fast and
lossless handover are described. Related works for
mainly fast and lossless handovers for Mobile IPv6
are discussed.
(Perez-Costa et al., 2003) study the performance
of Mobile IPv6, Hierarchical Mobile IPv6, Fast
Handover for Mobile IPv6 and their combination.
From their study, they show that the performance of
combination of HMIPv6 and FMIPv6 (H+F MIPv6)
is better than either HMIPv6 or FMIPv6. It has been
shown that it has better packet losses than FMIPv6
acting alone; but a larger bandwidth is obtained with
FMIPv6 acting alone. In case of H+F MIPv6, MAP
encapsulates all the data packets addressed to MNs,
and this overhead reduces the available bandwidth in
the channel. But for overall performance, H+F
MIPv6 has been found to be better.
(Lee et al., 2001) propose a scheme for fast and
lossless handover method considering DAD in IPv6-
based mobile/wireless networks. They have
considered the fact that latency comes mainly from
ND and DAD in stateless Auto-Configuration
scheme. MN obtains NCoA before handover and
PAR uses buffer management, making the proposed
scheme fast and lossless. In the proposed scheme,
MN receives several beacon signals containing the
network prefix and decides whether it needs to
change its AR by calculating the Received Signal
Strength (RSS) from neighboring ARs. Handover
initiates if another AR has higher signal strength
than the current AR. PAR starts buffering the
packets destined for MN. NAR, in the meantime,
acts as a proxy so that it can respond to any potential
DAD conflicts on its link for NCoA. NAR performs
a valid check for NCoA by comparing ND cache
entry with NCoA. PAR forwards the buffered
packets to NAR after receiving signal for the same
from NAR.
A scheme in which buffering of packets is done
at PAR while MN transitions to a new network is
proposed in (Park and Lim, 2002). Once MN
completes registration and obtains a valid NCoA,
PAR forwards the packets to MN at the new address.
ADVANCE DUPLICATE ADDRESS DETECTION (DAD) AND BUFFERING-BASED PROVISION OF SEAMLESS
HANDOVER IN MOBILE IPV6
45
Buffer management has been proposed with two
traffic classes, namely high-priority class as real-
time traffic with strict delay requirement such as
voice; and low-priority class with tolerable delay
and strict packet loss requirements such as pure data.
They propose an extension to the IPv6 Router
Advertisement which allows a router to advertise its
ability to support Simple Buffering (SB), where SB
is based on the general smooth handoff framework
as specified in (Krishnamurthi et al., 2001).
Different sub-options are also used; namely, Buffer
Initialize (BI), Buffer Forward (BF) and Buffer
Acknowledgement (BA). Incoming packets destined
for OCoA are buffered in addition to being
forwarded normally. When MN completes handover,
it sends BF sub-option asking the buffered packets
to be forwarded to NCoA.
A novel seamless handover architecture, S-MIP
is proposed in (Hsieh et al., 2003). The proposal
builds on top of hierarchical and fast-handover
mechanism of MIPv6, in conjunction with a
handover algorithm based on software-based
movement tracking techniques. They argue that with
such a combination, the performance is better,
providing a lossless handover with low latency. Two
distinct buffers are maintained at NAR in S-MIP
architecture, one for packets forwarded from PAR
(f-buffer) and one for the packets simulcast to the
current network MN is in and potential access
network MN will get attached to (s-buffer). NAR
will start delivering buffered packets to MN after it
receives Fast Neighbor Advertisement, signifying
that MN has arrived at its network. NAR will
attempt to transmit the packets in f-buffer and empty
it before beginning to transmit from s-buffer. At
PAR, it will only forward those packets which are
not simulcast to NAR. In case that MN does not
switch network immediately, it will therefore still be
able to receive packets from PAR.
A scheme called Mobile IPv6 Cache is proposed
in (Chung and Nelson, 2004). In the scheme, cache
at HA stores packets during handover. When HA
detects completion of handover, it will flush the
appropriate part of its cache immediately. Hence,
MN will receive a positive Binding
Acknowledgement followed by a burst of packets to
recover communications quickly. They argue that
Mobile IPv6 Cache can be implemented in existing
networks to improve communication performance
across handovers.
3 PROBLEM FORMULATION
From different handover schemes discussed, it
can be established that a combination of HMIPv6
and FMIPv6 has the advantage of reduction of
signaling delay and message overhead (as in
HMIPv6), and also support of fast handover by L2
triggers and minimal service disruption by tunneling
(as in FMIPv6). But a straight forward integration of
FMIPv6 with HMIPv6 would not be an efficient
option. Such integration would induce unnecessary
overhead for re-tunneling at the PARs and also
inefficient usage of network bandwidth as FMIPv6
uses the tunneling between the previous and new
ARs for fast handover. Hence, FHMIPv6, where
tunneling is between NAR and MAP rather than
between PAR and NAR, should perform better. MN
exchanges signaling messages for handover such as
RtSolPr, PrRtAdv, FBU, and FBAck with MAP. For
buffering, larger buffer at AR can tolerate less
frequent RAs and longer period of contact, but with
added latency. On the other hand, more frequent
RAs take up more wireless bandwidth and denser
coverage requires more ARs (hence, more
equipment). Balancing these factors is important for
achieving optimal handover.
The latency involved with Anticipated Fast
Handover protocol is mainly due to lack of
knowledge about NCoA, time taken for DAD,
signaling latency and L2 handover delay. The first
requires some time for discovery of new AR and for
BU. The second requires some time for the checking
of uniqueness of NCoA acquired by MN. The third
requires some time due to the signaling between
MN, ARs and HA / CN(s). The delays are usually
appreciable and can lead to packet loss. Effect of the
first case can be reduced to some extent by using
anticipation. Optimistic DAD (Moore, 2004), which
is a modification of existing IPv6 ND and Stateless
Address Auto-Configuration, could be one option to
reduce the effect of the second. Alternatively,
Address Pool based Stateful NCoA configuration for
FMIPv6 (Proactive and reactive stateful schemes)
could also be used (Jung et al., 2003). The scheme of
Advance DAD (Han et al., 2003) is another option.
Advance DAD scheme automatically allocates a
globally routable IPv6 CoA for the use of MNs that
participate in fast handover. Advance DAD is
considered in this work.
Buffering eliminates packet loss; they can reside
in ARs and store packets addressed to MN during
handover period temporarily. When handover is
complete and MN is attached to NAR, the stored
packets are forwarded to MN. Buffering could be
performed at NAR, in case of pre-registration, where
MN acquires NCoA before it moves to the new point
of attachment. Buffering could be performed at PAR
in case of post-registration. Tunnel Buffering (TB)
considers the situation where MN anticipates that it
is about to move, but does not know where it is
about to move to. In this case, MN can send a
ICETE 2005 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS
46
special L-BU requesting that its traffic be buffered at
MAP until it has determined its new link CoA
(Moore et al., 2004). Once it has completed
movement and gets attached to the new point of
attachment, MAP forwards the buffered packets to
MN and also starts tunneling the new packets
destined for MN.
In this work, buffering is performed at NAR for
FHMIPv6 protocol (Jung et al., 2004). MAP will
start forwarding packets destined to MN towards
NAR after tunnel is established during HI/HAck.
Hence, the packets are stored in the buffer at NAR
till MN establishes link with NAR, after which,
buffered packets are forwarded to MN.
3.1 Operational Detail
During normal operation, MN is connected to the
current AR, called PAR. Packets destined to MN are
routed from MAP via PAR in the wired link. In the
wireless link, the packets are sent on the air to MN
from PAR. A case of pre-registration handover is
considered. MAP manages a 'Passive Proxy Cache'
associated with its domain. The number of addresses
kept in the cache depends on the value of the address
pool.
The generation of globally routable addresses is
performed in the background and DAD is performed
on them. When DAD is complete, the addresses are
stored in the cache and are reserved. But when MAP
detects the use of any of the addresses present in the
cache, through RD messages, it deletes the address
from the cache. The address pool is continuously
updated by generating new addresses, so that the
addresses in the pool approach the pool size.
MN
(PAR)
MAP
PrRtAdv
F
-
BU
Disconnec
t
NAR
HI
HAc
k
MN
(NAR)
F
-
BAc
k
Buffering
BU
BAc
k
Deliver Packets
Normal HM IPv 6
Operation
Connec
t
NAc
k
No rmal
HMIPv6
Operation
PAR
NA
RA
Tu n n e
RtSolPr
Forward
Packets
Figure 1: Signalling and Operations during and after
Handover
To initiate fast handover, as a consequence of L2
handover anticipation trigger (Yegin et al., 2002),
MN sends Router Solicitation for Proxy (RtSolPr) to
MAP indicating that it desires to implement fast
handover to a new point of attachment. RtSolPr
contains an identifier for the new attachment point.
MAP will then send the Proxy Router
Advertisement (PrRtAdv) containing the duplication
free address from its cache and the address is
removed from the cache. Meanwhile, MAP sends
Handover Initiate (HI) message to NAR. In response
to the HI message from the MAP, NAR sends
handover acknowledgement (HAck) message and
setup of the tunnel between MAP and NAR is
complete. NAR starts buffering any packets tunneled
to it by MAP. After receiving PrRtAdv from PAR,
MN sends F-BU to MAP via PAR and disconnects
from PAR. MAP in response to F-BU associates
NCoA with OCoA and sends out acknowledgement
F-BAck, which MN receives from NAR. NAR
continues buffering the packets from MAP until it
receives Fast Neighbor Advertisement (F-NA)
message from the newly incoming MN after it has
established link connectivity. In response to F-NA
from MN, NAR sends out F-NAck to MN and starts
delivering the packets which have been buffered.
MN then follows the normal HMIPv6 operations by
sending Local Binding Update (L-BU) to MAP.
When MAP receives the new L-BU from MN, it will
clear the tunnel established for fast handover after
forwarding of the buffered packets is complete. In
response to L-BU, MAP will send Local Binding
Acknowledgement (L-BAck) to MN and the normal
operation of HMIPv6 will follow. The signaling and
operations are shown in figure 1 for the periods
during and after handover.
In (Hsieh et al., 2002), forwarding of packets by
PAR towards the NAR starts only after receiving
BU (F-BU) and in (Jung et al., 2004), forwarding
starts only after FBAck. But with these options,
some packets tend to go towards PAR and reach
PAR after MN has already left PAR. With such a
provision, it was seen that packets are lost. This will
be discussed a little later in the section on
Simulation. If the packets are tunneled towards NAR
right after HI/HAck, then this probability of loss is
decreased. The increase in latency with this is also
not very significant.
4 SIMULATION
Simulation was performed using Network
Simulator-2 (ns-2) (The Vint Project). The version
of NS-2 used was ns-2.1b7a, and the extension
developed by Jorg Widmer (Widmer) was added. In
ADVANCE DUPLICATE ADDRESS DETECTION (DAD) AND BUFFERING-BASED PROVISION OF SEAMLESS
HANDOVER IN MOBILE IPV6
47
addition, the extension for FHMIPv6 (Hsieh) was
also added on top of ns-2.1b7a and the NOAH
extension. The resulting product was modified as
required, the details of which are given in the
following.
The modifications consisted of changing the
point of sending different fast handover signals to
MAP instead of PAR. In the extension for FHMIPv6
developed by Hsieh, the handover signal RtSolPr
was sent to PAR by MN, HI and HAck were
communicated between PAR and NAR, and PAR
sent PrRtAdv to MN. Our case required sending
RtSolPr to MAP, HI and HAck to be exchanged
between MAP and NAR and PrRtAdv to be sent to
MN by MAP. Separate provisions for FBU and
FBAck were also added. MN sends FBU to MAP
from PAR and receives FBAck from NAR.
Tunneling process starting after HI/HAck was
also implemented, by which MAP tunnels packets
destined to MN towards NAR instead of PAR. In
addition, buffering option was also added. Buffering
was implemented in the ARs, where packets are
buffered until MN attaches itself to NAR. Classifiers
had to be modified for this. Classifiers sit inside a
node and they use the computed routing table to
perform packet forwarding. The buffer used is
limited by size, number of packets it can store, and
by the time limit for which the buffered packets are
acceptable. The time limit is the time period after
which, the packet loses its significance and hence is
discarded.
4.1 Simulation Parameters
The simulation model that was used is shown in
figure 2. The model used composed of HA, CN,
MAP, ARs and other routers.
CN and HA were connected via the Internet to
MAP. Router1 was included to simulate the
connection of MAP to HA and CN via the Internet
and a delay of 50ms from MAP and a bandwidth of
100Mbps was used. Since hierarchical structure was
assumed, the routers for different subnets were
connected to MAP.
All nodes below the MAP were members of the
same domain, hence the constant delay of 2 ms was
assumed for their links. The access routers, PAR and
NAR, representing different subnets were connected
via two different Intermediate Routers (Router2 and
Router3) to MAP. The wired links were modeled as
10Mbps duplex link with 2 ms delay and 1000Kbps
duplex link with 2 ms delay, from MAP to IRs and
from IRs to ARs respectively. For the wireless
medium, the LAN 802.11 access provided by ns-2
was used.
100Mbps
50ms
Router1
100Mbps
2ms
100Mbps
2ms
CN
(
0
)
HA
(
5
)
2Mbps IEEE 802.11 LAN 802.11 2Mbps IEEE 802.11 LAN 802.11
1000Kbps
2ms
1000Kbps
2ms
10 Mbps
2ms
10 Mbps
2ms
Router3 Router2
MAP
(1)
IR
(3)
IR
(
4
)
PAR
(7)
NAR
(8)
IR (2)
Mobile Node
Figure 2: The Simulation Scenario
The traffic source considered was CBR, with
UDP as the transport protocol with source at CN and
Null Agent at MN. The data packet was taken to be
of 512 bytes. The Cell sizes covered by the Access
Routers were taken to be of 100 meters diameter
with some area overlapping between them. Distance
between the Access Routers was taken to be 70
meters. In each simulation run, MN starts moving
towards NAR from PAR in a straight line at 10
seconds into the simulation. The values of speed and
source data rate were varied for different scenarios.
Three speeds were considered; 1m/s (approx.
4km/h), 15m/s (approx. 55km/h) and 25m/s
(90km/h) as the pedestrian, normal vehicle and high
vehicle speeds respectively. The performance during
handover for change in the source data rate and MN
speed was studied.
4.2 Simulation Scenarios
Various scenarios were considered for simulation;
variations of packet loss, end-to-end packet delay
and handover latency with CBR rate and speed of
MN were studied. Each simulation run lasted for 60
seconds for MN speed of 1m/s and 40 seconds for
other cases.
ICETE 2005 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS
48
4.3 Results and Discussions
Time (sec)
Packet Unique ID
12.3 12.35 12.4 12.45 12.5 12.55 12.6 12.65 12.7 12.75 12.
8
590
595
600
605
610
615
620
625
630
635
Basic MIP
FHMIP
Proposed with buffer
Figure 3: Packet Arrival for the Different Schemes
Figure 3 shows the packet arrival time for three
different schemes, base MIPv6, F-HMIPv6 studied
in (Hsieh et al., 2002), and the proposed scheme.
From the figure, it can be said that the proposed
scheme performs slightly better than that in (Hsieh
et al., 2002). In addition, the packets that are lost for
the scheme in (Hsieh et al., 2002) are not lost.
From simulation results, it is seen that a packet
(packet ID 608) reaches MAP at 12.484secs. In case
of the FHMIPv6 from (Hsieh et al., 2002), the
packet gets forwarded to PAR, and hence gets lost
because MN has already left the Previous Access
Network. In the proposed scheme, however, the
packet gets forwarded towards NAR and is buffered
till MN attaches itself to NAR. Similarly another
packet (packet 609) gets lost in the former case
because it gets forwarded towards NAR but MN has
not yet established link connectivity with NAR.
CBR Rate (Kbps)
No. of Packets Lost
0
100 200 300 400 500 600 700 800 9
0
1
2
3
4
5
6
7
8
MN Speed 1m/s
MN Speed 15m/s
MN Speed 25m/s
Figure 4: Number of Packets lost for Different Cases
CBR Rate (Kbps)
Delay (msec)
0 100 200 300 400 500 600 700 800 9
0
40
45
50
55
60
65
70
No Buffering, speed of MN 25m/s
Buffering Size 6, T imelimit 35 msecs, speed of M N 25m/s
Figure 5: Handover Latency with and without Buffering
The difference in handover latencies with and
without buffering widens with increasing data rate.
The highest difference between the two latencies is
for the highest data rate under consideration, for
which more number of packets need to be stored.
The difference is shown in figure 5.
But in the latter, it gets buffered and is sent
towards MN when it establishes link connectivity
with NAR. In case of Basic Mobile IPv6, a total of
14 packets get lost (packets 735-748) in the process
of handover, without the fast handover provision.
From simulation, it was also observed that the
number of packets lost in absence of buffering is
proportional to the data rate and almost independent
of MN speed, as the lost packet number is the same
for all the speeds, except for a few cases for lower
data rates, as seen from figure 4. After the initial
transient, the packet loss shows constant linear
relationship with the data rate for the three speeds
under consideration.
Figure 6 shows the handover latency variation
with data rate. Again, it also depends more on data
rate than on MN speed. But compared to end-to-end
delay, the dependence on speed is slightly more. The
trend of handover latency with data rate is on the
increment.
CBR Rate (Kbps)
Delay (msec)
0 100 200 300 400 500 600 700 800 9
0
35
40
45
50
55
60
65
70
75
Buffer Size 8, Timelimit 35 msecs, speed of MN 1 m/s
Buffering size 6, Timelimit 35 msecs, speed of MN 15 m/s
Buffering Size 6, Timelimit 35 msecs, speed of MN 25 m/s
Figure 6: Handover Latency Variation with Data Rate
ADVANCE DUPLICATE ADDRESS DETECTION (DAD) AND BUFFERING-BASED PROVISION OF SEAMLESS
HANDOVER IN MOBILE IPV6
49
CBR Rate (kbps)
Avg Packet Delay (ms)
100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 90
67.565
67.57
67.575
67.58
67.585
67.59
67.595
67.6
67.605
67.61
6
7.
61
5
Figure 7: End-to-end Packet Delay with Buffering
End-to-end packet delay, from CN to MN, also
has a direct relationship with data rate. However, it
is also less dependent on MN speed when compared
to data rate. The variation in the delay is almost
linear with the data rate, as evident from figure 7.
When a specific data rate was considered and the
variations of the different performance factors with
MN speed were studied, the earlier conclusion that
the factors were independent of MN speed was
verified. The variations are shown in figures 8 and 9.
Handover latency variation with MN speed is not
very significant, though some variations can be seen.
The independence of packet loss variation with MN
speed is more apparent than handover latency.
The end-to-end packet delay can also said to be
not too much dependent on MN speed, again
compared to the data rate. The variation does not
change significantly with MN speed. From figure 9,
it can be seen that initially the delay decreases with
increasing speed, but is almost constant when
saturation is reached for MN speed.
All of the performance parameters considered in
this work have been rounded off to the nearest three
digits in case of delays. Hence, these are of the order
of microseconds.
MN Speed (m/s)
Delay (ms)
1234567891011121314151617181920212223242
5
35
40
45
50
55
Figure 8: Handover Latency Variation with MN Speed
Speed of MN (m/s)
Avg Packet Delay (ms)
2 4 6 8 10 12 14 16 18 20 22 24 2
6
67.567
67.567
67.567
67.567
67.567
67.567
67.568
67.568
67.568
67.568
67.568
Figure 9: End-to-end Delay Variation with MN Speed
5 CONCLUSION
It can be concluded that if packets destined to MN
are forwarded towards NAR after HI/HAck, slightly
earlier than FBAck, the probability of loss of packets
is decreased, without increasing the delay
significantly.
The handover delay, end-to-end delay and loss of
packets are quite independent of MN speed
compared to data rate. In case of variable data rate,
these are more dependent on data rate than MN
speed. In case of a particular data rate also, there are
some variations, but the variations are not very
significant.
From the variations seen with MN speed and
data rate, the conclusion could be drawn that the
performance with VoIP applications, with voice
considered CBR traffic, is better than for multimedia
applications, with multimedia traffic being Variable
Bit Rate traffic.
From the simulations, it was also observed that
fast handover fails in some cases when the RtSolPr
sent by MN is not received by PAR. Hence, the cell
size, the transmission powers of Access Routers and
MN also play a significant role in the overall
performance. But these should be as small as
possible for obvious reasons. As an extension to this
work, performance with the variation in these
parameters and also with the ping-pong effect
considered could be studied.
REFERENCES
Bound J., Camey, M., Perkins, C. and Droms, R., 2001.
Dynamic Host Configuration Protocol for IPv6
(DHCPv6), IETF Internet Draft, draft-ietf-dhc-
dhcpv6-20.txt, October 2001.
ICETE 2005 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS
50
Castelluccia, C., 1998. A Hierarchical Mobile IPv6
Proposal, INRIA Technical Report No. 226, INRIA,
November 1998.
Chung, S.-L.J. and Nelson, R., (2004), “Improving Mobile
IPv6 Performance Across Handovers”, Thesis, WAND
Research Group, March 2004.
http://wand.cs.waikato.ac.nz/pubs/205/pdf/thesis.pdf
Han, Y., Choi, J., Jang, H. and Park, S., 2003. Advance
Duplicate Address Detection, IETF Internet Draft,
draft-han-mobileip-adad-01.txt, IETF, July 2003.
Hsieh, R., Seneviratne, A., Soliman, H. and El-Malki, K.,
2002. Performance Analysis on Hierarchical Mobile
IPv6 with Fast-Handoff over End-to-End TCP,
Proceedings of GLOBECOM, November 2002.
Hsieh, R., Zhou, Z. G. and Seneviratne, A., 2003. S-MIP:
A Seamless Handoff Architecture for Mobile IP,
Proceedings of INFOCOM 2003, April 2003.
Hsieh, R. (2003), NS-2 Extension for FHMIPv6,
http://mobqos.ee.unsw.edu.au/~robert/nsinstall.php>
Jung, H.Y., Koh, S. J., Soliman, H. and Kim, D. Y., 2003.
Address Pool based Stateful NCoA Configuration for
FMIPv6, IETF Internet Draft, draft-jung-mipshop-
stateful-fmipv6-00.txt, IETF, August 2003.
Jung, H.Y., Koh, S. J., Soliman, H., El-Malki, K. and
Hartwell, B., 2004. Fast Handover for Hierarchical
MIPv6 (F-HMIPv6), IETF Internet Draft, draft-jung-
mobileip-fastho-hmipv6-04.txt, IETF, June 2004.
Koodli, R., 2003. Ed., Fast Handovers for Mobile IPv6,
IETF Internet Draft, draft-ietf-mobileip-fast-mipv6-
08.txt, IETF, October 2003.
Krishnamurthi, G., Chalmers, R.C. and Perkins, C.E.,
2001. Buffer Management for Smooth Handovers in
IPv6, IETF Internet Draft, draft-krishnamurthi-
mobileip-buffer6-01.txt, IETF, March 2001.
Lee, C.-W., Lee, J.H., Joo, S.-B., and Kang, C.-H., 2001,
Fast and Lossless Handoff Method considering
Duplicate Address Detection in IPv6-based
Mobile/Wireless Network, Proceedings SPIE Volume
4586, Wireless and Mobile Communications, pp. 134-
143, October 2001.
Montavont, N., and Noel, T., 2003. Analysis and
Evaluation of Mobile IPv6 Handovers over Wireless
LAN, Mobile Networks and Applications (MONET),
Volume 8 Number 3, pp. 643 – 653, December 2003.
Moore, N. S., 2004. Optimistic Duplicate Address
Detection for IPv6, IETF Internet Draft, draft-ietf-
ipv6-optimistic-dad-02.txt, IETF, September 2004.
Moore, N.S., Choi, J. and Pentland, B., 2004. Tunnel
Buffering for Mobile IPv6, IETF Internet Draft, draft--
moore-mobopts-tunnel-buffering-00.txt, IETF, July
2004.
Pack, S., and Choi, Y., 2003. Performance Analysis of
Fast Handover in Mobile IPv6 Networks, Lecture
Notes in Computer Science (LNCS), Volume 2775, pp.
679-691, Springer-Verlag, September 2003.
Park, B.S. and Lim, C.S., (2002), A New Mobile IPv6
based Buffering Scheme in the All-IP Network,
The
2002 International Technical Conference on
Circuits/Systems, Computers and Communications,
July 2002.
Perez-Costa, X., Torrent-Moreno, M., and Hartenstein, H.,
2003. A performance comparison of Mobile IPv6,
Hierarchical Mobile IPv6, Fast Handovers for Mobile
IPv6 and their combination, ACM Mobile Computing
and Communications Review (MC2R), Volume 7,Issue
4, October 2003.
Thomas, S. and Narten, T., 1998. IPv6 Stateless Address
Autoconfiguration, IETF RFC 2462, December 1998.
Vint Project, The, Network Simulator (ns), version 2,
http://www.isi.edu/nsnam/ns
Widmer, NOAH Extension for ns-2.1b7a,
http://www.icsi.berkeley.edu/~widmer/mnav/ns-
extension/
Yegin, A.E., Funato, D., El Malki, K., Gwon, Y., Kempf,
J., Pettersson, M., Roberts, P., Soliman, H. and
Takeshhita, A., 2002. Supporting Optimized Handover
for IP Mobility - Requirements for Underlying
Systems, draft-manyfolks-l2-mobilereq-02.txt, IETF
Internet Draft, IETF, June 2002.
ADVANCE DUPLICATE ADDRESS DETECTION (DAD) AND BUFFERING-BASED PROVISION OF SEAMLESS
HANDOVER IN MOBILE IPV6
51