FAST MOBILE IPV6 APPROACH FOR WIRELESS LAN

BASED NETWORKS

Link-Layer Triggering Support for IEEE 802.11

Norbert JORDAN and Alexander POROPATICH

Institute of Broadband Communications, Vienna University of Technology

Favoritenstrasse 9/388, A-1040 Vienna, Austria

Keywords: Mobile IPv6, Fast MIPv6, Wireless LAN, Link-Layer Triggering, Fast IEEE 802.11 Handover.

Abstract: The standard Mobile IPv6 specification provides comprehensive mobility management for the IPv6 proto-

col. During the handover there is a period in which the mobile node is unable to send or receive packets due

to link-layer switching and IPv6 protocol layer operations. This overall handoff latency resulting from base-

line MIPv6 procedures, namely movement detection, new care-of address configuration, and binding up-

dates with peer entities, is often unacceptable for any kind of real-time service (video-conferencing, voice-

over-IP,…). A new fast handover approach, based on Fast Handovers for Mobile IPv6, is proposed in this

paper, which will support seamless movement in between IPv6 domains using a IEEE 802.11 network infra-

structure. A new low latency handoff method for IEEE 802.11 will be proposed, where access point beacons

are utilized for carrying IPv6 prefix information without altering the Mobile IP or IEEE 802.11 specifica-

tions. A WLAN service will continuously monitor the radio signal quality of the attached access point and,

if necessary, will switch to another access point in range. This feature and the elimination of firmware-based

active scanning during link-layer handovers have the flavor effect of reducing the overall link-layer handoff

delay to about 10%. We will further introduce our wireless testbed infrastructure for evaluation of the pro-

posed approach. Performance evaluation is used to verify the effectiveness of our implementation and an

extensive simulative comparison is used for scalability analyses.

1 INTRODUCTION

Owing to the assistance of Mobile IPv6 (Johnson,

2004), a mobile node can effectively maintain its IP-

layer connectivity to the Internet when it changes its

point-of-attachment somewhere in the world. During

the accomplishment of the handover, the mobile

node is unable to send or receive IPv6 packets be-

cause of its L2 and also L3 handover operations.

This high handover latency is unacceptable to real-

time applications or delay sensitive traffic. Each

time a mobile client moves, it is necessary to per-

form movement detection by discovering (sending

router solicitation) its current point of attachment.

In Mobile IPv6 (Johnson, 2004), the movement de-

tection algorithm relies on the periodic sending of

router advertisements in order to enable the mobile

node to determine its current location. The only way

to improve the detection performance is to broadcast

router advertisements at a faster rate, which may

result in a poor link utilization. For that reason the

fast handover protocol (Koodli, 2004) is designed to

achieve a seamless handoff when mobile nodes

move from one domain to another.

In a mobile-initiated and anticipated fast-

handover scenario described in (Koodli, 2004), the

mobile node first sends a Router Solicitation for

Proxy (RtSolPr) message to the current access router

containing any Access Point specific identifiers. The

current Access Router replies with a Proxy Router

Advertisement (PrRtrAdv) message , which may

contain a subnet-specific information tuple [AP-ID,

AR-MAC, AR-IP]. This message exchange allows a

mobile node to obtain the new Access Router's pre-

fix information, which is needed to perform an “an-

ticipative” configuration of the new IPv6 address on

the new subnet. Figure 1 presents a general mobile-

initiated “predictive” fast handover scenario.

102

Jordan N. and Poropatich A. (2004).

FAST MOBILE IPV6 APPROACH FOR WIRELESS LAN BASED NETWORKS - Link-Layer Triggering Support for IEEE 802.11.

In Proceedings of the First International Conference on E-Business and Telecommunication Networks, pages 102-108

DOI: 10.5220/0001395301020108

 SciTePress

Figure 1: Reference Scenario for FMIPv6 Handover.

With the information provided in the PrRtAdv

message, the MN formulates a prospective new CoA

and sends a Fast Binding Update (FBU) message.

The purpose of FBU is to authorize the old AR to

bind the current Care-of address (CoA) to new CoA,

so that arriving packets can be tunneled to the new

location. Depending on whether an FBack (Fast

Binding Acknowledgement) is received prior to the

Mobile Node’s movement or not, the prospective

address can be used immediately after attaching to

the new subnet link. In case it moves without receiv-

ing an FBack, the MN can still start using the new

CoA after announcing its attachment through a Fast

Neighbor Advertisement (FNA) message (see Figure

2).

Figure 2: Message Flow for Mobile-Initiated HO.

However, the above protocol assumes that the

L2 protocol is capable of delivering the L2 identifier

of the new access point to the mobile node. More

important, to initiate a seamless handover, is the fact

that the current AR must be capable of mapping this

new L2 identifier into the IP address of the target

AR. We will show that all these requirements for

Fast MIPv6 can be fulfilled in our implementation

without any modifications to the IEEE 802.11 stan-

dards.

2 FAST HANDOVER FOR IEEE

802.11

The growing popularity of IEEE 802.11 (IEEE,

1999) has made “wireless” LAN a potential candi-

date technology for providing high speed reliable

wireless access services. In addition by supporting

Mobile IP, wireless LAN can meet demands for ex-

panded wireless access coverage while maintaining

continuous connectivity from one domain into an-

other. In order to be able to accomplish a fast hand-

over on Layer 3 it is necessary to implement a trig-

gered information indicated by the underlying link-

layer driver.

2.1 Link-Layer Triggering

In order to achieve an efficient interworking be-

tween Fast Mobile IPv6 and IEEE 802.11, it is nec-

essary that the link-layer initiates the handover. The

mobile node normally does this by sending a proxy

router solicitation at the IP layer. This action is trig-

gered by the underlying link layer in the mobile

node, which must be aware that a handover is about

to take place. This is the only possible way since

from the IEEE 802.11 link-layer’s point of view the

mobile node is the only entity which is aware, that

the host is about to attach to a new AP. In our im-

plementation there is a tool running at the mobile

node which continuously monitors the signal

strength of the attached AP. In case the receiving

power-level falls below a pre-defined value, the tool

reacts by collecting information of all APs in range.

So the tool is able to anticipate the best destination

for the handover. At the same time of preparing the

link-layer handover to the most qualified AP, the

client-tool will send a trigger message to the fast-

handover module. The next step that follows is the

proposed FMIPv6 approach explained in Section 3.

2.2 Enhanced WLAN Handover

Even if the Fast Mobile IP approach is implemented

properly, there are still delay issues to solve during

the link-layer handover. Since Mobile IP and link-

layer handover should go hand-in-hand, there is still

an unsolved problem with the Layer 2 handoff-

latency when the mobile node moves from one AP

to another. There exists a definite period of time in

which the mobile node is unreachable due to the

FAST MOBILE IPV6 APPROACH FOR WIRELESS LAN BASED NETWORKS - Link-Layer Triggering Support for

IEEE 802.11

103

layer 2 movement (i.e. re-synchronization with the

new AP). It has to be remarked that the exact

amount of time varies, depending on the deployed

WLAN technology. Some measurements (Velayos,

2003) (Velayos, 2004) (Mishra, 2002) for IEEE

802.11b show that this time period can vary from

200 to 1500 ms, depending on the type of vendor

equipment.

Figure 3: IEEE 802.11b HO Latency without Optimi-

zation.

The main problem during the handover is the

fact that stations have to detect the lack of radio

connectivity based on unsuccessful frame transmis-

sions. The difficulty is to determine the reason for

the failure among collision, radio signal fading or

the station being out of range. In our implementation

the signal strength is monitored continuously. In

case that the signaling level drops below a prede-

fined threshold, the tool automatically tries to hand-

off to an AP in range, which provides a much better

connectivity. So the long phase of detection can be

saved and the handover is carried out much faster.

This WLAN handover-tool takes advantage of the

information provided by the physical layer and com-

pletely skips the detection phase. Stations equipped

with our tool start the search phase when the quality

of the radio-signal falls below a pre-defined thresh-

old. Therefore, the search always starts before any

frame has been lost. This has the favorable effect

that the overall handover-time can be reduced to

about 350ms, as demonstrated in (Jordan, 2003).

Another issue of WLAN is the active-scan process,

which is often enforced with each AP-handover.

Preventing active-scanning, additionally helps to

reduce the link-layer latency to about 60 to 100 ms

(depending on vendor hardware).

Figure 4: Optimized IEEE 802.11b Handover.

These initial improvements will enable wireless

networks to carry real-time applications along the

infrastructure.

3 PROPOSED FAST HANDOVER

APPROACH

As already stated in Section 2 our implementation

helps the Mobile Node to detect if the current link is

degrading and therefore starts searching for a new

AP with improved link-quality. To do this, the mo-

bile node scans all possible frequencies (specified by

the IEEE 802.11b standard) [10] and compares the

received signal with the one currently received. If

the mobile node finds a better signal it can switch to

the new AP. But the mobile node’s link layer im-

plementation does not know whether this AP is at-

tached to a new AR. The link layer only knows

about link layer addresses and the AP’s SSID (Ser-

vice Set Identifier) string. However, if the AP

name/link layer address (which identifies an AP) is

known, the mobile node’s IP-layer implementation

can request that the current AR should provide the

prefix/router address, which the new AP is attached

to. This idea assumes that an AR is configured with

a table containing its own and the neighboring APs

link-layer addresses and their corresponding AR.

In our implementation we configure each access

point involved with a special SSID string (e.g.: SSID

= “2001:200:8:72AB:1434::1/64”) which further

implicitly presents all information about the prefix

of the attached AR. Whenever the mobile node an-

ticipates a handoff, the handover-tool exactly knows

the prefix of the new AR the AP is attached to. In

that way the mobile node performs “anticipative”

configuration of the new IP address on the new sub-

net using the router prefix information carried in the

beacon message of the new AP. If more than one

destination access point is in range, the mobile node

100

120

140

160

180

600 650 700 750 800 850 900 950

Handoff-Latency (ms)

Occurrence

100

125

150

175

200

225

250

50 60 70 80 90 100

Handoff-Latency (ms)

Occurrence

ICETE 2004 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS

104

could prefer to carry out a movement to an AP

within the same subnet. Thus only a link-layer hand-

over would be performed, which further improves

the handoff-latency in this special case. In all other

cases the mobile node will perform the configuration

of a new IP address and continues with the Fast Mo-

bile IPv6 handover until the mobile node arrives at

the new AR (NAR).

4 IMPLEMENTATION OVER-

VIEW

To make a serious network evaluation in the area of

Mobile IPv6 possible, we implemented an enhanced

IPv6 testbed which is connected to the worldwide

native “6net” infrastructure. As it can be seen in

Figure 5, we built up a central core network where

all subnetworks are attached to. In between each

included network provider, we implemented WAN-

Emulators that thwart all IPv6 packets transmitted.

As our major aim was to create a very flexible net-

work infrastructure, we put a single WAN-Emulator

for each provider. So we are able to tune the link-

delay individually, depending on the appropriate

scenarios to be analyzed. Wireless LAN IEEE

802.11b and IEEE 802.11g are deployed in the over-

all infrastructure.

Figure 5: Mobile IP Testbed at TU-Vienna.

Three independent network operator domains

were deployed, whereas one includes the Home

Agent for the mobile node experiments. Further-

more, another network operator domain includes

some kind of hierarchical structure in order to be

able to do a performance comparison with the alter-

native HMIPv6 approach. All hosts including mobile

nodes, correspondent nodes and the routers within

each provider’s area have RedHat Linux 8.0 in-

stalled with Kernel 2.4.22. For the MIPv6 basis

functionality we utilized MIPL 1.0, provided by

Helsinki University of Technology (HUT).

The Linux driver for all WLAN activities is

based on the HostAP project, which seems to be the

most flexible environment for making link-layer

triggering realizable in a very fast manner. HostAP

provides a general Linux driver for all

PRISM2/2.5/3 based Wireless LAN cards. The re-

sults of an initial link-layer trigger optimization can

be seen in Figure 3 and 4. These measurements are

deployed by skipping the active-scanning mecha-

nism within each handover.

5 PERFORMANCE EVALUATION

In this section we present initial results for a verifi-

cation of the implemented IPv6 mechanisms and

furthermore results based on our real-world Mobile

IPv6 network infrastructure. For all measurements

we derived average-values from about 1000 samples

for each point in the graphics. This helps us to get

significant and serious results for comparing of

standard Mobile IPv6 to the enhanced FMIPv6 ap-

proach.

The first graph presents the difference in be-

tween communication with and without Route Op-

timization. The results of the end-to-end delay, de-

pending on various link-delays, are presented in Fig-

ure 6.

Figure 6: Route Optimization Impact on End-to-End De-

lay.

Figure 7 depicts the dependence of the handoff

latency (foreign link – foreign link) on the variance

of sending Router Advertisements. Obviously, the

handoff latency falls off as Router Advertisement

messages are sent more frequently.

0,5

1,5

2,5

3,5

1 10 100 1000

Link Delay (ms)

End-to-End Delay (s)_

MIPv6

no Route Optimization

FAST MOBILE IPV6 APPROACH FOR WIRELESS LAN BASED NETWORKS - Link-Layer Triggering Support for

IEEE 802.11

105

Figure 7: Handoff Latency for varying Router-Adv. Inter-

val.

The results in Figure 8 and Figure 9 present the

average handoff latency with dependence on the

link-delay between different networks. Here we di-

rectly compared basic Mobile IP to the Fast MIPv6

approach.

Figure 8: Average Handoff-Delay for Basic Mobile IPv6.

Figure 9: Average Handoff-Delay for Fast Mobile IPv6.

As already assumed from the Fast Mobile IPv6

approach, the packet loss during a handover between

different network providers is decreased to a mini-

mum compared to basic Mobile IPv6. Figure 10 de-

picts the packet loss results for an Iperf- generated

UDP-data stream of 160 kbit/s in between the mo-

bile node and its correspondent node. As illustrated

in Figure 5 the Correspondent Node is placed near

the core network.

Figure 10: Average Number of Packet Loss during HO.

6 SIMULATIVE COMPARISON

For a deeper understanding as well as for a more

general evaluation of Mobile IPv6 in an environment

with many users, the use of simulations is indispen-

sable. We performed a simulative comparison of

baseline Mobile IPv6 and the Fast Handoff approach

in an wireless LAN based scenario, comprising 4

independent operator domains with 10 home users

per access router. Even if the focus is on the evalua-

tion of MIPv6 bases protocols, we also include the

impact of a shared-link environment based on IEEE

802.11b.

6.1 Simulation Scenario

For the performance study of MIPv6 we decided to

evaluate a basic scenario which is simple enough to

get results in a reasonable time but also complex

enough to get an expressive feeling for real-world

provider scenarios. The studied scenario (see Figure

11) is composed of a group of Correspondent Nodes,

one for each Mobile Node, connected to one central

router (CR) through the IPv6 backbone. Each access

router (AR) represents a different IP subnet and acts

as a home agent for 10 mobile nodes. All Mobile

Nodes are located at their home link when the simu-

lation starts. Either the distance in between the ARs

and also the transmitted signal-power are chosen in a

way to create overlapping coverage areas for ena-

bling seamless movement in between the various

domains (see Figure 12).

0,01 0,1 1 10

Radvd-Interval (s)

Average Handoff-Latency (s) _

FL-FL (20 -30)

1 10 100 1000

Link Delay (ms)

Average Handoff-Latency (s)_

FL-FL (20 -30 )

HL-FL (10-20)

1 10 100 1000

Link Delay (ms)

Average Handoff-Latency (s)_

FMIPv6 FL-FL

FMIPv6 HL-FL

1 10 100 1000

Link Delay (ms)

Average packet loss per handoff

Basic MIPv6

Fast MIPv6

ICETE 2004 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS

106

Figure 11: Mobile IPv6 – Simulation Scenario.

The random way-point mobility model is used

for all Mobile Nodes, which is best suited for realis-

tic user movement. Connectivity for each Mobile

Node is provided by IEEE 802.11 using 2 Mbit/s

and DCF and traffic is assumed to be UDP with 40

kbit/s constant bit rate.

Figure 12: Access Router Range Topology.

For all our experiments we used the ns-2 (Ns2,

1998) simulation tool, whereas the MOBIWAN add-

on by Thierry Ernst (Ernst, 2002) is deployed to get

basic MIPv6 functionality into the simulator. Further

essential MIPv6-specific software code was adopted

from a MIPv6 simulation environment by NEC

Europe in Germany.

6.2 FMIPv6 Simulation Results

In this section, we present the results of our ns-2

simulative comparison of baseline Mobile IPv6 and

the enhanced Fast Handover mechanism.

Figure 13 presents the comparison of the handoff

latency obtained during basic Mobile IPv6 handoff

with the latency resulting from a Fast Mobile IPv6

handover. The simulation results show that similar

to the performance measurements in Section 5 we

also achieve some latency-related advantage for sce-

narios with a huge number of concurrent moving

users.

Figure 13: Average Handoff Latency Comparison.

With a reduced latency also the packet loss dur-

ing the handoff can be reduced consequentially for

the Fast Handoff approach. This behavior, similar to

our testbed results from Section 5, is demonstrated

in Figure 14.

Figure 14: Average Packet-Loss per Handover.

7 CONCLUSIONS AND FUTURE

WORK

With this work we presented a first evaluation and

simulative results of a Fast Mobile IPv6 handover

approach for wireless LAN based networks. Our

evaluation showed that a client based fast handover

approach can be suitable to improve WLAN hand-

overs for real-time traffic and enables better mobility

management support in IEEE 802.11 based wireless

LANs. In the near future we will investigate on hier-

archical approaches for IPv6 and other smart solu-

tions with improved handoff latency performance

and reduced signaling overhead.

0,5

1,5

2,5

3,5

4,5

1 10 100 1000

Link Delay (ms)

Average Handoff-Latency (s)_

Basic MIPv6 HL-FL

FMIPv6 HL-FL

1 10 100 1000

Link Delay (ms)

Average packet loss per handoff

Basic MIPv6

Fas t MIPv6

FAST MOBILE IPV6 APPROACH FOR WIRELESS LAN BASED NETWORKS - Link-Layer Triggering Support for

IEEE 802.11

107

ACKNOWLEDGEMENTS

Part of this work has been performed within the pro-

ject "WISQY - Wireless InterSystem Quality-of-

Service" at the Telecommunications Research Cen-

ter Vienna (ftw) and has been funded in the frame-

work of the Austrian Kplus Competence Center Pro-

gramme. We would also like to thank Uschi Christa-

lon-te Kock from NETGEAR for the great support

within our research and Richard Menedetter for his

assistance during the measurements.

REFERENCES

Proceedings Papers:

Johnson, D.B., Perkins, C.E., and Arkko, J., 2004. Mobil-

ity Support for IPv6, RFC 3775, IETF Network Work-

ing Group.

Koodli, R., 2004. Fast Handovers for Mobile IPv6, Inter-

net-Draft, IETF Mobile IP Working Group, < draft-

ietf-mipshop-fast-mipv6-02.txt>.

IEEE, 1999. Part 11: Wireless LAN Medium Access Con-

trol (MAC) and Physical Layer (PHY) Specifications,

IEEE Standard 802.11.

Velayos, H., and Karlsson, G., 2003. Techniques to Re-

duce IEEE 802.11b MAC Layer Handover Time, KTH

Technical Report, ISSN 1651-7717, Stockholm, Swe-

den.

Velayos, H., and Karlsson, G., 2004. Techniques to Re-

duce IEEE 802.11b Handoff Time, , IEEE ICC, Paris,

France.

Mishra, A., Shin, M., and Arbaugh, W., 2002. An Empiri-

cal Analysis of the IEEE 802.11 MAC Layer, Techni-

cal Report CS-TR-4395, University of Maryland, De-

partment of Computer Science.

Jordan, N., Fleck, R., and Ploninger, C., 2003. Fast Hand-

over Support in Wireless LAN based Networks, Fifth

IFIP-TC6 International Conference on Mobile and

Wireless Communications Networks, Singapore, Sin-

gapore.

Ns2, 1998. Network Simulator (ns), ). version 2 , UC

Berkeley, http://www.isi.edu/nsnam/ns.

Ernst, T., 2002. Mobiwan: ns-2 extensions to study mobil-

ity in Wide-Area IPv6 Networks, MOTOROLA Labs

Paris, http://www.inrialpes.fr/planete/mobiwan.

Costa, X.P., Schmitz, R., Hartenstein, H., and Liebsch, M.,

2002. A MIPv6, FMIPv6 and HMIPv6 Handover La-

tency Study Analytical Approach, Network Laborato-

ries, NEC Europe Ltd.

Books:

Gast, M. S., 2002. 802.11 Wireless Networks, The

Definitive Guide. Sebastopol, CA: O'Reilly & Asso-

ciates, Inc.

Intersil Corporation, 2001. PRISM Driver Programmer's

Manual, Development Information, Version 2.10,

(For Distribution Under NDA Only).

ICETE 2004 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS

108