Simplified fast handover in mobile IPv6 networks (2)

This paper proposes an enhancement to the FMIPv6, the Simplified Fast Handover in Mobile IPv6 Networks SFMIPv6, which significantly reduces the antic-ipation time of the fast handover an

Simplified fast handover in mobile IPv6 networks

Kongju National University, 182 Sinkwan-dong, Gongju-si, Chungcheongnam 314-701, Republic of Korea

a r t i c l e i n f o

Article history:

Available online 24 June 2008

Keywords:

Mobile IPv6

Fast handover

Simplified fast handover

a b s t r a c t

The Fast Handovers for Mobile IPv6 (FMIPv6) protocol provides seamless handover; it uses anticipation based on layer 2 trigger information of the mobile node (MN) to obtain a new care-of address at the new link while still connected to the previous link, thus reducing handover delay A bidirectional tunnel is then established between access routers to minimize packet loss during the handover However, this method incurs higher signaling costs compared with the standard Mobile IPv6 protocol In many cases, the mobile node cannot complete the fast handover in predictive mode due to lack of time, especially with high-speed movement of the mobile node This paper proposes an enhancement to the FMIPv6, the Simplified Fast Handover in Mobile IPv6 Networks (SFMIPv6), which significantly reduces the antic-ipation time of the fast handover and thereby increases the probability that the protocol can perform the fast handover in predictive mode In this paper, we also present performance evaluations in terms of the influence of two factors, the break-down point and the velocity of the MN, using evaluation models The numerical results prove that the network performance of the proposed protocol is effectively improved compared to the original protocols, the FMIPv6 and Fast Handover Support in Hierarchical Mobile IPv6 (F-HMIPv6) Moreover, the results show that the proposed protocol could appropriately operate with high-speed mobile node movement

Ó2008 Elsevier B.V All rights reserved

1 Introduction

Mobile IPv6 (MIPv6)[1]is a global mobility management

proto-col in mobile IPv6 networks that specifies the operations through

which a mobile node (MN) maintains its connectivity to the

inter-net while being handed over from one access router to another

These operations cause considerable and unacceptable packet

losses and latencies in certain circumstances such as real-time

applications Several extensions to Mobile IPv6 have been

pro-posed in order to reduce these handover latencies and packet

losses

The FMIPv6[2]is an enhancement to Mobile IPv6, proposed by

the Internet Engineering Task Force (IETF) that provides seamless

handover in Mobile IPv6 Networks Layer 2 (L2) trigger information

from the mobile node (MN) is used to obtain a valid new care-of

address (NCoA) while it is still connected to the previous link,

and then a bidirectional tunnel is established between the previous

access router (PAR) and the new access router (NAR) in order to

re-duce packet loss during the handover The MN can use the NCoA

immediately after establishing a connection with the new link;

thus, the handover latency and packet loss can be considerably

re-duced Furthermore, if the MN has enough time to perform fast

handover in the predictive mode, there may be no packet loss

However, a certain cost is also incurred in terms of additional over-head; for example, new signaling messages are required for the anticipated handover procedure Therefore, in many cases the

MN cannot successfully perform the fast handover procedure in predictive mode due to lack of time In these cases, the MN must revert to the normal MIPv6 or switch to the reactive mode of the FMIPv6, depending on the link-break time; both handover latency and packet loss are thereby increased Several extensions[7–12] have been proposed to improve the performance of the FMIPv6, but these studies did not consider reducing the anticipated over delay that limits the time for the MN to perform the fast hand-over procedure in predictive mode Morehand-over, all of these enhancements issue more signaling messages during this critical period; therefore these proposals are inappropriate for high-speed

MN movement

This paper proposes a modification to the FMIPv6 that consider-ably reduces the overhead associated with fast handover, including the signaling cost and the packet delivery cost This method may significantly increase the probability that the MN can successfully perform the fast handover procedure in predictive mode, thus sup-porting high-speed MN movement

The key idea is to eliminate the useless time during which the

MN waits for the PAR’s response after sending a Router Solicitation for Proxy (RoSolPr) message, before completely sending the Fast Binding Update (F-BU) message to the PAR Instead, the MN uses the F-BU option within the RtSolPr message, such that the PAR can initiate fast handover immediately after receiving the RtSolPr

0140-3664/$ - see front matter Ó 2008 Elsevier B.V All rights reserved.

* Corresponding author Tel.: +82 10 2891 7299.

E-mail addresses: nvhanh@kongju.ac.kr (N Van Hanh), rosh@kongju.ac.kr (S Ro),

arkil@kongju.ac.kr (J Ryu).

Contents lists available atScienceDirect

Computer Communications

j o u r n a l h o m e p a g e : w w w e l s e v i e r c o m / l o c a t e / c o m c o m

Accordingly, the anticipation time of the fast handover can be

sig-nificantly reduced about a roundtrip time (RTT) on the wireless

link between the MN and the PAR

For the sake of simplicity, we describe the proposed scheme

based on the FMIPv6[2], but it can also be applied to other schemes

that support fast handover, such as the Fast Handover for

Hierarchi-cal Mobile IPv6 (F-HMIPv6)[4] However, in numerical results, we

show the comparison between the original schemes (FMIPv6 and

F-HMIPv6) and the proposed schemes (SFMIPv6 and SF-HMIPv6 that

apply to the respective original schemes, FMIPv6 and F-HMIPv6) in

order to illustrate the improvement of the new proposal

The rest of the paper is organized as follows Section2provides

the background and related work Then, in Section3, we describe

our proposed protocol, the Simplified Fast Handover in Mobile

IPv6 Networks (SFMIPv6) Scheme In Section4, the performance

of the proposed protocol is estimated through evaluation models

Numerical results are given in Section5 Next, the security issues

of the proposal protocol are discussed in Section6 Finally, we

con-clude this paper in Section7

2 Background and related work

2.1 Fast Handovers for Mobile IPv6

In this section, we briefly introduce the FMIPv6 Detailed

infor-mation can be found in[2] The basic operation of the FMIPv6 is

illustrated inFig 1(a)

The FMIPv6 introduces seven additional message types: Router

Solicitation for Proxy Advertisement (RtSolPr), Proxy Router

Adver-tisement (PrRrAdv), Handover Initiate (HI), Handover

Acknowledgement BAck), and Fast Neighbor Advertisement

(F-NA)

A fast handover procedure starts with the MN sending an

RtSolPr message, and ends with the MN receiving an F-BAck

mes-sage on the previous link In the FMIPv6 protocol, when an MN is

aware of its movement towards an NAR through an L2 trigger,

the MN must perform a fast handover procedure Then, after

con-necting to the NAR, the MN immediately sends an F-NA message

without the need for route discovery in order to inform its

pres-ence, so that arriving and buffered packets can be forwarded to

the MN

Finally, as in the MIPv6 protocol, in order to complete the hand-over, the MN must perform home registration with the Home Agent (HA) and correspondent registration, including a return rou-tability procedure and binding update with the CN

2.2 Predictive and reactive modes of the FMIPv6

If the MN receives the F-BAck message on the previous link be-fore connecting to the new link, it is in the predictive mode of fast handover However, since the FMIPv6 depends on the layer 2 trig-ger, there is no assurance that the MN has enough time to initiate and complete the fast handover procedure while it is still con-nected to the previous link If connectivity to the previous link is lost unexpectedly, the MN has to operate in reactive mode and one of the following cases can occur:

1 RtSolPr message cannot be sent or PrRtAdv message cannot be received over the previous link: In this case, the MN has no information about the new link, anticipation fails, and the MN reverts to the standard Mobile IPv6 handover

2 F-BU message cannot be sent over the previous link: In this case, the PAR cannot initiate fast handover with the NAR; there-fore the NCoA cannot be validated, and the PAR cannot buffer and forward packets to the new link Accordingly, packets that are en route to the PCoA of the MN may be lost

3 F-BAck cannot be received over the previous link: In this case, the MN sent the F-BU but did not receive F-BAck before the link break The PAR redirects packets to the NAR, which buffers the packets until the MN announces its presence The MN may pos-sibly receive F-BAck on the new link

From[5,6], the probability, Ploss, of an MN losing its connection with the previous link during the fast handover procedure, is given by

Ploss¼ 1  e x tL2Trigger

ð1Þ

where – tL2Triggeris the L2 trigger time taken from the occurrence of an L2 trigger event to the link break-down point;

– xis a decreasing factor which is introduced to account for a variety of decreasing patterns

The link break-down point usually depends on the size of

over-lap area and also the velocity of the MN

2.3 Extensions to the FMIPv6 protocol

Several extensions[7–12]have been proposed to improve the

performance of the FMIPv6, but these studies did not consider

reducing the anticipated handover delay that limits the time for

the MN to perform the fast handover procedure in predictive

mode In addition, all of these enhancements issue more signaling

messages during this critical period; therefore these proposals are

inappropriate for high-speed MN movement

Malki and Soliman[7]introduced the Fast Handover protocol

with a simultaneous binding function to minimize packet loss at

the MN To reduce packet losses, a mobile node’s traffic is bi-casted

or n-casted to all locations where the MN could roam next in the

near future, as well as to its current location This procedure

elim-inates any ambiguity regarding the moment at which the traffic

should be rerouted toward the mobile node’s new location after

a fast handover, and it enables the protocol to decouple L2 and

L3 handovers However, Simultaneous Binding still retains all the

steps of the fast handover procedure; thus the probability that

the MN cannot perform the fast handover in predictive mode is still

high, as in normal FMIPv6

Leoleis et al.[8]proposed an integrated unicast and multicast

handover support solution to the FMIPv6, Seamless Multicast

Handover It is approached in a twofold manner: first by enabling

the new access router to become a recipient of the multicast traffic

of interest via tunneling, and second, by buffering the tunneled

traffic for the period during which the mobile is unable to

commu-nicate due to link layer communication unavailability Tunneled

multicast packets are de-capsulated before being natively

for-warded on the wireless link, eliminating the overhead caused from

packet encapsulation over the air interface However, signaling

messages are increased even more, and the probability of the MN

performing the fast handover in predictive mode cannot be

improved

Chen and Zhang[9]presented a modification to the FMIPv6

pro-tocol using extra binding updates to reduce tunneling time

be-tween the PAR and NAR Pre-Binding Update and Pre-Binding

Acknowledgement messages are exchanged between the MN and

the Home Agent (HA)/Correspondent Node (CN) before the Fast

Binding Update message is sent to the PAR Thus, the reverse

tun-nel between the PAR and the NAR need not be established

How-ever, other issues could arise First, the Pre-Binding Update

message is sent with the unverified NCoA to the CN Thus, if the

NCoA does not then pass the Duplicate Address Detection (DAD)

test, these binding update messages would be useless Second,

since two more new signaling messages are issued during the

crit-ical time of the fast handover procedure, this method could

pro-long the fast handover delay, easily leading to fast handover

failure

Zhang and Pearce[10]proposed that the care-of address test, a

part of the Return Routability test for Mobile IPv6 route

optimiza-tion, be run proactively in the context of the FMIPv6 protocol so

that the latency caused by the care-of address test after move-ments can be reduced The key idea is to deliver the Care-of Test Init (CoTI) message as soon as possible from the MN to the NAR Once the NCoA is determined by the NAR, the NAR modifies (if nec-essary) and forwards the CoTI message to the CN to launch the care-of address test There is no improvement for the fast handover procedure Furthermore, the CoTI message is encapsulated in the fast binding update message, so it can incur even higher signaling costs than the original protocol

Kim and Kim[11] introduced an early binding fast handover (EBFH), in which an MN performs an early fast binding update with its current access router before a trigger that signals that an MN is closed to handover The FMIPv6 initiates movement detection through a link-going-down trigger, whereas EBFH completes its binding update for the NCoA before the link-going-down trigger The purpose of EBFH is to provide a fast handover for fast-moving nodes If the MN moves at high speed, it reverts to the FMIPv6 This requires that the MN detect the speed and direction of movement; the implementation of such a detection mechanism is outside the scope of this proposal Furthermore, EBFH issues many signaling messages before the link-going-down trigger, so it consumes a large amount of network performance and creates significant use-less overhead

Hsieh et al.[12]presented an architecture which enhances the integrated hierarchical and fast handover scheme in conjunction with a handover algorithm based on a pure software-based move-ment tracking technique This scheme could minimize handover latency and virtually eliminate packet loss at the layer 3 IP layer However, this scheme introduces a new agent, the Decision Engine (DE), and six more new messages Thus, it consumes a large amount of network performance and creates significant useless overhead Moreover, too many messages are exchanged during the critical time of the fast handover procedure, prolonging the anticipation time of the fast handover and easily leading to failure

of the fast handover in predictive mode The supposed advantages

of this scheme therefore cannot be achieved

3 The proposed protocol The proposed scheme, SFMIPv6, is an optional and fully back-ward-compatible enhancement to the FMIPv6 protocol

In the FMIPv6, after sending the RtSolPr message, the MN must wait for a PrRtAdv message from the PAR before obtaining the NAR’s information to generate a new CoA The PAR must also wait for an F-BU message from the MN in order to initiate fast handover Such delays are unnecessary and could be eliminated

Fig 1(b) shows the basic operation of the SFMIPv6 scheme (in comparison with the FMIPv6 scheme inFig 1(a)) The SFMIPv6 scheme does not define any new messages A bit, M, and a new op-tion, F-BU Opop-tion, are introduced as shown inFig 2(b) Bit M is used within the RtSolPr and PrRtAdv messages to instruct the receiving node to operate in the SFMIPv6 scheme

Upon receiving the F-BU Option within the RtSolPr message, the PAR forms a new CoA on behalf of the MN Immediately afterward,

Reserved

LLA Options

F-BU Option

Sequence #

Sub-options

the PAR initiates the fast handover; to the PAR does not wait for the

F-BU message from the MN after it sending the PrRtAdv message

The basic operation of the protocol in predictive mode

(Fig 1(b)) is as follows:

– Upon receiving an indication from a wireless link-layer trigger,

the MN initiates the fast handover procedure by sending an

RtSolPr message containing the F-BU option and the M bit to

the PAR

– After receiving the RtSolPr, the PAR checks whether the M bit is

set in order to follow the SFMIPv6 scheme If the M bit is set,

the PAR forms a prospective new CoA on behalf of the MN, then

simultaneously sends an HI message to the NAR, and the

PrR-tAdv with M bit set to the MN

– According to the M bit within the PrRtAdv received by the MN,

it will decide to follow the FMIPv6 scheme or the SFMIPv6 If

the M bit is set, it indicates that the PAR supports the SFMIPv6

protocol, so the MN does not need to send the F-BU to the PAR

It just needs to wait for an F-BAck

– Upon receiving the HI message, as specified in the FMIPv6,

the NAR validates the uniqueness of the NCoA that was

already formed by the PAR and is included in the HI

mes-sage The NAR then responds to the PAR with a HAck

message

– After receiving the HAck message from the NAR, the PAR sends

an FBAck message to the MN on both links, the previous link

and the new link, and starts the tunneling of buffered and

arriving data toward the NCoA

– As soon as it is connected to the new link, the MN sends an FNA

message to inform the NAR of its presence Packets are

deliv-ered to the MN from this point on

– In order to complete the handover, as in the MIPv6 and the

FMIPv6 protocols, the MN must perform a home registration

with the Home Agent (HA) and correspondent registration

including return routability procedure and binding update

with the CN Then new packets can be sent directly from the

CN to the NCoA of the MN

If the PAR does not support SMIPv6, the M bit in the RtSolPr and

PrRtAdv messages and the F-BU option within the RtSolPr message

would be ignored and have no effect All normal FMIPv6 messages [2]remain unchanged and retain their original meaning

The proposed scheme can also be applied to other schemes that support fast handover, such as the Fast Handover for Hierarchical Mobile IPv6 (F-HMIPv6) and called SF-HMIPv6 F-HMIPv6is the combination of Fast Handovers for Mobile IPv6 (FMIPv6)[2]and Hierarchical Mobile IPv6 Mobility Management (HMIPv6)[3] The basic operation of SF-HMIPv6 is similar to the SFMIPv6, but the

MN exchanges the signaling messages for the handover with MAP, rather than with PAR as the F-HMIPv6

4 Performance evaluation Since the fast handover procedure depends on the layer 2 trig-ger, in some cases the MN may not have enough time to initiate and complete the fast handover while still connected to the previ-ous link In this paper, we evaluate the performance in terms of the impacts of two factors, the break-down point and the velocity of the MN To evaluate the impact of the break-down point and the velocity of the MN on the network performance, the evaluation model is as follows

4.1 Evaluation model 4.1.1 Coverage model For the sake of simplicity, we consider a wireless network sys-tem in which access points (AP) are used with omni-direction antennas organized in a hexagonal pattern as shown in Fig 3, and placed at the center of cells with uniform topographical and lo-cal signal propagation conditions Fig 3(a) shows the layout of overlapping coverage areas in the system The coverage area can

be defined in terms of Signal Strength; the effective coverage is the area in which MNs can establish a link with acceptable signal quality with the AP This area can be modeled by a circle centered

at the corresponding AP The coverage radius, r, is defined as the distance from an AP to its coverage boundary The cell radius, c,

is the distance from an AP to its cell boundary.[14]shows four dif-ferent coverage models based on the ratio of the coverage radius to the cell radius (c/r) In this paper, we use a model as shown in Fig 3 There are three overlapping coverage areas shown in

A

E

B

G

F

C

D

Effective Coverage Cell Boundary

r c

d

a

b

Fig 3(b), denoted by I, II, and III Area I is a non-overlapping region

in which the MN can receive signals from only one AP, while both

areas II and III are overlapping regions in which the MN is able to

receive from different APs: two APs in area II, and three APs in area

III

4.1.2 The minimum overlapping distance

In order to evaluate the influence of the velocity of the MN on

the network performance, we consider the worst case, in which

the traveling path of the MN through the overlapping area is the

minimum overlapping distance.Fig 4shows the minimum

over-lapping distance and related notations that are defined above

From[12], A0B0is a side of an internal equilateral triangle that is

formed by overlapping area III (inFig 4)

The distance between APs, d, is given by

d ¼ c ffiffiffi

3

p

ð2Þ

To find the minimum overlapping distance, x (A0B0), as a function of

r and c, considerDBPA Using trigonometric relationships we obtain

cosðbÞ ¼BP

2

þ BA2 ðPAÞ2

2BP  BA or cosðbÞ ¼

r2þ 3c2 x2

2rc ffiffiffi 3

sinðbÞ ¼PHBP or x ¼ 2r  sinðbÞ ð4Þ

sin2ðbÞ ¼ 1  cos2ðbÞ ð5Þ

From(3)–(5), we obtain

cosðbÞ ¼

ffiffiffi

3

p

4r c þpffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi4r2 3c2

ð6Þ

Applying(4)–(6), the minimum overlapping distance, x, is given by

x ¼ 2r  sin cos1

ffiffiffi 3 p 4rc þpffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi4r2 3c2

!!

ð7Þ

4.1.3 Handover scenario

We assume that there is no change in direction while the MN

moves inside the overlapping area

Fig 5shows the basic handover scenario The optimal handover

point occurs at position A However, the relative signal strength

with hysteresis margin is a commonly used approach to prevent

the ping-ponging phenomenon[15] In this approach, a handover

is only triggered if the signal level of the AP to which the MN is cur-rently attached differs from one of new APs by at least the hyster-esis margin h, as at position B The associated delay dh of the hysteresis margin h is introduced in[13], and is given by

dh¼2vd 1 þ 10

Applying(8), the distance dhfrom the optimal handover point (po-sition A) to po(po-sition B given by

dh¼ vdh¼d21 þ 10

4.1.4 Network topology model

In order to evaluate the network performance, we use the sys-tem model shown inFig 6 There are six nodes participating in the system model, and each node is connected to the others by a link Each link is assigned a notation associated with the latency

or cost of a packet delivery between two nodes of the link The spe-cific notations are described inTable 1

4.2 Influence of break-down point

If connectivity to the previous link is lost unexpectedly, net-work performance could degrade depending on the link-break time.Fig 7shows the timeline for the fast handover procedure,

A

B

C

Effective Coverage Cell Boundary

r

β

H

c

O B Q

x

Minimum Overlapping Distance

' '

h

MN

Handover Triggered Optimal

Handover Point

Signal Strength

dn

y x z

Fig 5 Basic handover scenario.

Fig 8shows the timeline for failure cases, andFig 9shows the

comparison of handover latency among FMIPv6/F-HMIPv6 with/

without the proposed protocol In this section, we evaluate

perfor-mance while varying the link break-down point, which is related to

the failure cases of the fast handover We then compare the

origi-nal FMIPv6 protocol with the proposed protocol

T specified inTables 2 and 3and the following is given by T = X,

or T = min{z/v, D0} depending on the context of evaluation, the break-down point, or the velocity of the MN

[6]classified failure cases in the FMIPv6 into three cases as de-scribed in Section2.2 In the SFMIPv6, we also classify failure cases into three cases and evaluate network performance for each case, different from those in the FMIPv6, as follows:

– Failure Case 1 (X < m): In this case, the MN recognizes the link-break with the PAR before the PAR completely receives the RtSolPr message The MN has no information about the new link, and the NAR has no information about the MN Thus, the MN must switch to the standard MIPv6 protocol.Having

no information about the new link, the MN has to perform movement detection As the discussion in[18], the movement detection delay (DMD) of the MIPv6 is given by DMD= DRA/2 Then, upon receiving the new network prefix advertised from the NAR, the MN configures an NCoA and performs DAD pro-cess for this NCoA, i.e DIP= DDAD+ DMD Since the fast handover was not inititated, no packets are bufferred, i.e Cbuffer= 0 Accordingly, all of packets destined to the MN during the dis-ruption time are lost, i.e Closs= k(D+ DL2+ DIP+ DBU+ Dnew) – Failure Case 2 (m 6 X < 2m): In this case, the MN recognizes the link-break with the PAR before receiving the PrRtAdv message Although the MN has no information about the new link, the PAR can still initiate fast handover.Having no information about the new link, the MN has to perform movement detec-tion As the discussion in[18], the movement detection delay (DMD) is given by DMD= DRA/2 However, the MN performs the fast handover in reactive mode, and thus DIP= DMD+ 2(m + 2n) Packets arrived at PAR after the reception of the HAck message (i.e at the time of m + 4n) are buffered until the the MN compeletes the conrespondent registration and receives the first packet directly from the CN, i.e Cbuffer= k{T +D+ DL2+

DIP+ DBU+ Dnew (m + 4n)} If the link break-down point before the buffering time (i.e m + 4n), packets destined to the

MN during in the interval [X, m + 4n] are lost In addition, pack-ets still in the wireless link (from the PAR to the MN) while the link is broken are also lost, i.e Closs= max{0, k(m + 4n  T)} + km

– Failure Case 3 (2m 6 X < 2m + 4n): In this case, the MN recog-nizes the link-break with the PAR before the MN receives F-BAck message However, the MN has information about the new link before the link break-down point according to the reception of a PrRtAdv message (with M bit set) Subsequently, immediately after connecting to the new link, the MN needs not to perform movement detection process, i.e DMD= 0 The

DIP= 2(m + 2n) As described in failure case 2, C buf-fer= k{T +D+ DL2+ DIP+ DBU+ Dnew (m + 4n)}, and Closs= max {0, k(m + 4n  T)} + km

In successful case (X P 2m + 4n), no packets are buffered and lost due to the handover Immedidately after connecting to the new link, the MN sends a F-NA massage to the NAR to inform its presence and receive packets destined to the MN via previous path Then, after the completion of the correspondent registration, the

MN can receive packets directly from the CN Thus, DMD= 0,

DIP= 2m, Cbuffer= 0, Closs= 0

The summary of the above evaluations is shown inTable 3

In this paper, we compare network performance between the FMIPv6, F-HMIPv6 schemes and the proposed enhancements to those original schemes, SFMIPv6 and SF-HMIPv6, respectively For the sake of simplicity, in this section we evaluate the perfor-mance of only the FMIPv6 and SMIPv6 The others can easily be de-rived from the evaluations inTables 2 and 3

3m+2n 3m+4n 4m+4n

MN sends

RtSolPr

MN receives PrRtAdv

PAR sends

PrRtAdv

PAR receives F-BU NAR receives HI

PAR receives HAck

MN receives F-BAck

2m+4n

MN sends

RtSolPr

PAR sends

PrRtAdv/HI

MN receives F-BAck

2m

MN receives PrRtAdv

Table 1

List of evaluation parameters

Symbol Description

m Latency or cost of a packet delivery between MN and Access Router (PAR or

NAR)

n Latency or cost of a packet delivery between Intermediate Router (IR)/

Mobility Anchor Point (MAP) and Access Router (PAR or NAR)

a Latency or cost of a packet delivery between HA and IR/MAP

b Latency or cost of a packet delivery between CN and IR/MAP

X The exact time point at which the MN loses its connectivity with the PAR

D The additional time taken for the MN to recognize the link-break with the

PAR after the link has actually been broken

k Packet arrival rate

D 0 The time span between the moment the MN sends the RtSolPr message to

the PAR/MAP and the moment the F-BAck arrives at the MN (in theory) The

value of D 0 depends on protocols It is (4m + 4n), (2m + 4n), (4m + 6n), or

(2m + 4n) in FMIPv6, SFMIPv6, F-HMIPv6, or SF-HMIPv6 protocols,

respectively

D L2 Layer 2 handover delay

D L3 Layer 3 handover delay: the time span between the moment the MN

completes layer 2 handover and receives the first packet from CN

D IP IP connectivity delay, which depends on each scheme and the failure or

success of the fast handover

D DAD Duplicated Address Detection Delay

D RA The mean value of the Router Advertisement intervals

D MD The Movement Detection Delay in standard Mobile IPv6

D MN-HA One way transmission delay between MN and HA

D MN-CN One way transmission delay between MN and CN

D HA-CN One way transmission delay between HA and CN

D BU Binding update delay with HA and CN in Mobile IPv6

D new D new = D MN-CN

D total Overall handover latency

C buffer Buffering Cost: total number of buffering packets at the PAR during the fast

handover

C loss Loss Cost: total number of lost packets during the fast handover

r The coverage radius in coverage model

c The cell radius in coverage model

d The distance between APs in coverage model

x The minimum overlapping distance

v The velocity of the MN

h Hysteresis margin, representing the signal level about which the new AP is

stronger than the old one.

K 2 K 2 represents environment-specific attenuation characteristics [13]

d h The associated delay d h of the hysteresis margin h

d h The distance from the optimal handover point ( Fig 5 ) to the hysteresis

margin point (position B), which is associated with d h

z The remaining distance in the overlapping area of the MN from the

hysteresis margin point

From[6], the packet delivery cost for failure, Cpacket, can be

mea-sured by

wheresis the weighting factor

The total packet delivery cost Ctotal(t) for a fast handover can be measured by

where Ploss(t) is the probability that an MN can lose its connection with the previous link at time t as shown in(1) In this case, we con-sider the link break-down time as a variable and it is a unique var-iable in the equation

4.3 Influence of velocity

In this section, we evaluate the performance while varying the velocity of the MN To evaluate the influence of the velocity of the MN on network performance, we consider the worst case, in which the traveling path of the MN through the overlapping area

is the minimum overlapping distance, x Let

From(9) and (12), we obtain the position of the MN in the overlap-ping area, y, which is given by

y ¼x2 þd21 þ 10

Thus, the remaining distance in the overlapping area of the MN for handing off, z, is given by

z ¼ x  y ¼2 x d21 þ 10

1 þ 10h=K 2 ð14Þ

MN sends RtSolPr

Link break-down point

CoA Configured

L3 handover delay(DL3) Link break

recognized

Receive Packet from CN

Link break recognition delay

L2 handover delay (DL2)

IP connectivity delay (DI)

Binding Updates Completed

DBU Dnew

X+ +DL2+DL3

Link Up

Fig 8 Timeline for failure cases in fast handover.

-BU

HAck

PrRtAdv

HI

F-BAck

MRtSolPr

HAck

MPrRtAdv

HI

F-BAck

NAR

PAR

MN

NAR

PAR

MN

2m+2n

4m+2n

HAck

PrRtAdv

HI

F-BAck

MRtSolPr

HAck

MPrRtAdv

6 HI

F-BAck

NAR

MAP

MN

NAR

MAP

MN

2m+4n 4m+4n

Fig 9 Comparison of handover latency among the FMIPv6/F-HMIPv6 with/without

the proposed protocol.

Table 2

Performance evaluation of FMIPv6 scheme

Velocity of MN v > z/2m z/3m < v 6 z/2m z/(4m + 4n) < v 6 z/3m v 6 z/(4m + 4n)

D total T + D + D L2 + D IP + D BU + D new T + D + D L2 + D IP + D BU + D new T + D + D L2 + D IP + D BU + D new T + D L2 + D IP + D BU + D new

C buffer 0 0 k{T + D + D L2 + D IP + D BU + D new  (3m + 4n)} 0

C loss k(D + D L2 + D IP + D BU + D new ) k(D + D L2 + D IP + D BU + D new ) max{0, k(3m + 4n  T)} + km 0

Table 3

Performance evaluation of SFMIPv6 scheme

Velocity of MN v > z/m z/2m < v 6 z/m z/(2m + 4n) < v 6 z/2m v 6 z/(2m + 4n)

D total T + D + D L2 + D IP + D BU + D new T + D + D L2 + D IP + D BU + D new T + D + D L2 + D IP + D BU + D new T + D L2 + D IP + D BU + D new

C buffer 0 k{T + D + D L2 + D IP + D BU + D new  (m + 4n)} k{T + D + D L2 + D IP + D BU + D new  (m + 4n)} 0

C loss k(D + D L2 + D IP + D BU + D new ) max{0, k(m + 4n  T)} + km max{0, k(m + 4n  T)} + km 0

Fig 5shows the representation of x, y, and z in the context of the

handover scenario

Let V denote the velocity of an MN in a cell, which is a

contin-uous random variable in the interval [0, Vmax], where Vmax> 0 is

gi-ven Let us assume that the velocity is distributed uniformly in the

interval [0, Vmax] From [17], the CDF (Cumulative Distribution

Function) of velocity is given by

FVðvÞ ¼

0; if v < 0

v

V max; if 0 6 v 6 Vmax

1; if v > Vmax

8

>

>

ð15Þ

The probability, PS(v), of an MN successfully performing the fast

handover procedure in predictive mode is given by

PSðvÞ ¼ P v <Dz

0

¼





z D0

0

v

Vmax¼D z

Thus, the probability, PF(v), of an MN failing to successfully perform

the fast handover procedure in predictive mode is given by

PFðvÞ ¼ P v >Dz

0

¼ 1 D z

Therefore, the overall packet delivery cost for failure can be given by

The evaluation analysis of the influence of velocity on the network

performance of the FMIPv6 and the proposed protocol, SFMIPv6,

based on z, are evaluated and shown inTables 2 and 3

5 Numerical results

The parameter values used in the numerical analysis are shown

in Table 4 based on discussions in [6,13,16,18] Among these

parameters, the packet arrival rate has been chosen for a high

traf-fic volume with k = 2 packets/ms (i.e 2000 packets/s)

5.1 Influence of link break-down point

In this section, we compare the network performance of the

FMIP, SFMIPv6, F-HMIPv6, and SF-HMIPv6 protocols in terms of

the impact of the velocity of the MN

Fig 10shows and compares the handover latencies of the

pro-tocols while varying the link break-down point X, which is

illus-trated in Fig 8 When the MN loses its connectivity with the previous link before receiving the PrRtAdv massage, it has no information about the NAR and must switch to the standard MIPv6 Thus, the total handover latencies of the protocols are high

In this case, the handover latencies of the original and proposed protocols are the same However, if the link is broken after the

MN has completed sending the PrRtAdv message (at 6 ms for SFMIPv6; at 8 ms for SF-HMIPv6), the handover latencies of the proposed protocols are greatly reduced compared to the original protocols

Fig 11shows the comparison of the packet loss cost of the pro-tocols while varying the link break-down point The packet loss costs of the proposed protocols, SFMIPv6 and SF-HMIPv6, are much lower than those of the original protocols Indeed, in the SFMIPv6/ SF-HMIPv6, the PAR can initiate the fast handover with the NAR immediately after receiving the RtSolPr message (at 6 ms and

8 ms in the SFMIPv6 and SF-HMIPv6, respectively), therefore the packet loss cost is very small or zero from 6 ms/8 ms However,

in the FMIPv6/F-HMIPv6, the PAR can initiate the fast handover with the PAR only after receiving the F-BU message, three times la-ter than in the SFMIPv6/SF-HMIPv6 (at 18 ms/24 ms, respectively) Consequently, the packet loss cost of the SFMIPv6/SF-HMIPv6 is greatly reduced as shown

Fig 12shows the comparison of the total packet delivery cost of the protocols while varying the L2 trigger time The packet delivery cost of the proposed protocols is much lower than that of the ori-ginal protocols

Table 4

Parameters used in numerical analysis

Fig 11 Packet loss cost in terms of break-down point.

5.2 Influence of velocity of MN

Based on discussions in [13,14], we consider two scenarios

depending on the coverage radius (r) and cell radius (c) in order

to evaluate the network performance in terms of the influence of

the velocity of the MN as follows:

– Scenario 1: r = 60 m and c = 40 m

– Scenario 2: r = 30 m and c = 20 m

Fig 13presents the effect of the velocity of the MN on the

hand-over latencies of the protocols and compares the performance

be-tween these protocols InFig 13(a), r and c are fixed at 60 m and

40 m, respectively; in Fig 13(b), r and c are set to 30 m and

20 m, respectively In the context ofFig 13(a), the proposed

proto-cols, SFMIPv6 and SF-HMIPv6, have enough time to initiate fast

handover for MN velocities up to 180 km/h, whereas the F-HMIPv6

and FMIPv6 have to switch to the standard Mobile IPv6 when the

speed reaches 125 km/h (F-HMIPv6) and 170 km/h (FMIPv6) In the context ofFig 13(b), the overlapping area is decreased by half, and the traveling distance is also decreased by half Thus, the im-pact of the velocity on the handover latency of the protocols is sig-nificantly increased due to lack of time to complete the fast handover, but it is much lower for the proposed protocols Indeed, the protocols must revert to the standard Mobile IPv6 when the speed of the F-HMIPv6, FMIPv6, SF-HMIPv6, and SMIPv6 reach

60 km/h, 85 km/h, 125 km/h, and 170 km/h, respectively Thus, in this context, the limits of the velocity of the MN in the proposed protocols, such that the handover latency can be acceptable, are twice as high as those in the original protocols

Fig 14shows the packet loss cost of the protocols as a function

of the velocity of the MN In the context ofFig 14(a) (r = 60 m and

c = 40 m) the proposed protocols have enough time to perform the fast handover procedure, so the packet loss cost is zero or very small However, the original protocols must revert to the standard Mobile IPv6 when the speed of the MN reaches 125 km/h (FMIPv6) and 170 km/h (F-HMIPv6) Thus, the packet loss costs of these pro-tocols are very high In the context ofFig 14(b), the coverage and cell radii are decreased by half, so the traveling distance of the MN

in the overlapping area is also decreased by half In this case, the

MN does not have sufficient time to perform the fast handover when the MN moves at high speed However, in the proposed pro-tocols, the effect of the speed of the MN on the packet loss cost is much lower than that in the original protocols

Fig 15shows the comparison of the total packet delivery cost between the protocols while varying the velocity of the MN In the context ofFig 15(a) (r = 60 m and c = 40 m) the speed of the

MN is not so influential in the total packet delivery cost of the pro-posed protocols, whereas it significantly affects the total packet delivery cost of the original protocols Moreover, in the context

ofFig 15(b), when r and c are decreased by half, the impact of

Fig 12 Total packet delivery cost in terms of L2 triggering time.

Fig 13 Handover latency in terms of velocity.

the speed of the MN is much higher on the overall packet delivery

cost of the original protocols, whereas the proposed protocols

could effectively operate at much higher MN speeds, up to

125 km (SF-HMIPv6) and 170 km/h (SFMIPv6)

In summary, the performances of the proposed protocols are

effectively improved compared to the original protocols It has

been shown that our proposal could appropriately operate with

high-speed MN movement

6 Security discussion

The proposal does not introduce new security vulnerabilities

than those already described in the FMIPv6[1] However, the

pro-posal does not use the F-BU message; instead the F-BU option is

in-cluded in the RtSolPr message Therefore, the RtSolPr message in

the proposed scheme must be secured and authenticated as the

F-BU message in the FMIPv6

7 Conclusion

This paper proposes an enhancement to the FMIPv6 which

sig-nificantly reduces the overhead associated with fast handover,

including the signaling cost and packet delivery cost, by

optimiz-ing the fast handover procedure The key point is to eliminate the

useless time during which the MN waits for the response of the

PAR after sending a RoSolPr message, before completely sending

an F-BU message to the PAR Instead, the MN uses the F-BU

op-tion within the RtSolPr message, such that the PAR can initiate

fast handover immediately after receiving the RtSolPr

Accord-ingly, the anticipation time can be remarkably reduced about a

RTT on the wireless link between the MN and the PAR Therefore,

this method can significantly increase the probability that the MN

successfully performs the fast handover procedure in predictive

mode

In this paper, we considered evaluation models in order to

eval-uate and compare network performance between the original

pro-tocols, the FMIPv6 and F-HMIPv6 propro-tocols, and the proposed

protocols (that apply to the respective original protocols), SFMIPv6

and SF-HMIPv6, while varying the link break-down point and the

velocity of the mobile node The numerical results show that the

network performances of the proposed protocols are effectively

improved compared to the original protocols Moreover, the

influ-ence of the velocity of the MN on the network performances of the

protocols was evaluated, and the evaluation results showed that

the proposed protocol could appropriately operate in the presence

of high-speed mobile node movement

References

[1] D Johnson, C Perkin, J Arkko, Mobility Support in IPv6, IETF, RFC 3775, June 2004.

[2] R Koodli, Fast Handovers for Mobile IPv6, IETF, RFC 4068, July 2005 [3] H Soliman, C Castelluccia, K El Malki, L Bellier, Hierarchical Mobile IPv6 Mobility Management, IETF, RFC 4140, August 2005.

[4] HeeYoung, Koh SeokJoo, Fast handover support in hierarchical Mobile IPv6, in: The 6th International Conference on Advanced Communication Technology, vol 2, 2004, pp 551–554.

[5] Sangheon Pack, Yanghee Choi, Performance analysis of fast handover in mobile IPv6 networks, in: Proceedings of IFIP Personal Wireless Communications (PWC) 2003, Venice, Italy, September 2003.

[6] Dong Su, Sang-Jo Yoo, Fast handover failure-case analysis in hierarchical mobile IPv6 networks, IEICE Transactions on Communications (2005) [7] El K Malki, H Soliman, Simultaneous bindings for mobile IPv6 fast handovers, IETF, draft-elmalki-mobileip-bicasting-v6-06-txt, July 2005.

[8] Georgios A Leoleis, George N Prezerakos, Iakovos S Venieris, Seamless multicast mobility support using fast MIPv6 extensions, Computer Communications 29 (18) (2006) 3745–3765.

[9] HongFei Chen, Jian Zhang, Prep-binding of fast handovers for mobile IPv6, draft-chen-mipshop-fast-handovers-prep-binding-02.txt, IETF, April 2006 [10] J Zhang, D Pearce, Proactive care-of address test for route optimization in FMIPv6, draft-zhang-mipshop-proactive-cot-00.txt, IETF, August 9, 2005 [11] Hocheal Kim, Youngtak Kim, An early binding fast handover for high-speed mobile nodes on MIPv6 over connectionless packet radio link, in: Seventh ACIS International Conference on Software Engineering, Artificial Intelligence, Networking, and Parallel/Distributed Computing (SNPD’06), 2006, pp 237–242 [12] R Hsieh, Z.G Zhou, A Seneviratne, S-MIP: a seamless handoff architecture for mobile IP, INFOCOM 2003, Twenty-second Annual Joint Conference of the IEEE Computer and Communications Societies, IEEE, vol 3, 30 March–3 April 2003,

pp 1774–1784.

[13] M Emmelmann, Influence of velocity on the handover delay associated with a radio-signal-measurement-based handover decision, Technical Report TKN-05-003, Telecommunication Networks Group, Technische Universität Berlin, Berlin, Germany, April 2005 (online) Available from: <http://www.tkn tu-berlin.de /publications/papers/TKN-Report-05-003.pdf>.

[14] N Srivastava, S.S Rappaport, Models for overlapping coverage areas in cellular and micro-cellular communication systems, Global Telecommunications Conference, 1991, GLOBECOM’91, Countdown to the New Millennium, Featuring a Mini-Theme on: Personal Communications Services, vol 2, 2–5 December 1991, pp 890–894.

[15] Qing-An Zeng, Dharma P Agrawal, Handoff in wireless mobile networks, Handbook of Wireless Networks and Mobile Computing, John Wiley & Sons, Inc., 2002 pp 1–25.

[16] Jane-Hwa Huang, Li-Chun Wang, Chung-Ju Chang, Deployment strategies of access points for outdoor wireless local area networks, VTC 2005-Spring, 2005 IEEE 61st, vol 5, 30 May–1 June 2005, pp 2949–2953

[17] M Cho, K Kim, E Szidarovszky, Y You, K Cho, Numerical analysis of the dwell time distribution in mobile cellular communication systems, IEICE Transactions on Communications E81-B (4) (1998) 715–721.

[18] L JunSeob, M JaeHong, K SangHa, Considerations for designing fast handoff mechanisms in mobile IPv6, in: The 6th International Conference on Advanced Communication Technology, vol 1, 2004, pp 21–24.

Fig 15 Total packet delivery cost in terms of velocity.