Performance Analysis on IP- based Soft Handover across

ALL-IP Wireless Networks

Farouk Belghoul, Yan Moret, Christian Bonnet

Department of Mobile Communications, Institute Eurecom,

2229 Route des Crêtes, BP 193, F-06904 Sophia Antipolis Cédex, France

Abstract. Mobile users are facing the fact that many heterogenous radio access

technologies coexist, ranging from wireless LANs to cellular systems. No technology has

emerged as common and universal solution which makes the current trends today toward

design of All-IP wireless networks, where radio cells are under the control of IP Access

Routers for signalling and data transmission. In such as networks, an IP-device with

multiple radio interfaces or software radio can roam between different radio networks

regardless the heterogeneity of radio access technologies. The design of an All-IP wireless

network requires an efficient and flexible IP-based handover management, and a major

issue in handover control is how to reduce data loss and avoid additional end to end

transmission delay. In this paper we propose and evaluate mechanisms to handle soft-

handover management in IP layer over heterogenous networks. Those mechanisms coexist

with Mobile IPv6 and allow efficient micro mobility management.

Keywords. All-IP wireless networks, heterogeneous network, Soft Handoff, Mobile IP,

IEEE 802.11

1 Introduction

One of the most important issues in IP-mobility protocols design is the IP handover

performance. IP handover occurs when a mobile node changes its network point of

attachment from an Old Access Router (OAR) to a New Access Router (NAR). If not

performed efficiently, end-to-end transmission delay, jitters and packet loss directly

impact and disrupt applications perceived quality of services. Because Soft handover

provides same data receiving from multiples Access Router, it allows mobile station’s

session to progress without interruption when a Mobile Node (MN) moves from one radio

cell to another. These can be done, if and only if 1.MN is able to communicate

simultaneously with multiple ARs in the same time. 2. The network can duplicate and

Belghoul F., Moret Y. and Bonnet C. (2004).

Performance Analysis on IP- based Soft Handover across ALL-IP Wireless Networks.

In Proceedings of the 1st International Workshop on Ubiquitous Computing, pages 83-93

DOI: 10.5220/0002686500830093

 SciTePress

correctly merge the IP-flows from the correspondent node (CN) to the MN through

different access routers. If the two conditions are verified, it is possible to eliminate

packet loss and reduces end-to-end transmission delays, which provides a clear advantage

to traffic requiring real time transmission. This paper presents pure IPv6 Soft Handover

mechanisms [2], based on IPv6 flows duplication and merging in order to offer pure IP-

based mobility management over heterogenous networks. Proposed solution does not

impose any change to the Mobile IPv6 standard [3]. It is an extension to support an

efficient Soft handover and micro mobility management, for Mobile Node (MN) with

multiple radio interfaces (WLAN) [4] or with unique CDMA interface. This solution

requires the introduction of new component called “Duplication & Merging Agent”

(D&M) agent. It is a conventional router located at the core network used to duplicate and

merge IPv6 flows to and from the MN.

2 Related Works

Mobile IPv4 and Mobile IPv6 [3][5] introduce basic mobility management services in

Internet Protocol, their simplicity and scalability give them a growing success. MIP poor

Handover performance makes it not appropriate for real time applications with heavy

constraints in transmission delays and packets loss. Smooth handover [6] introduces

packets buffering mechanisms in each access router to recover all lost packets during

handover, but introduces additional end-to-end packets transmission delays. Basic MIP

Fast Handover [7] is another approach that anticipates the obtention and the registration of

future mobile address. “MIP fast handover bi-casting”[8] exploits this anticipation to

simultaneously duplicate data to the old and new care of address (CoA) of MN, which

allows MN to receive data immediately after performing layer 2 handover and removes

layer 3 handover delays. Simulation work done in [9][10] , shows us that globally,

smooth and fast handover can not avoid TCP performances degradation and UDP packets

loss when MN moves from OAR to the NAR . In the following we propose a novel IP

based handover scheme that addresses delays and packets loss issues in the same time and

across heterogeneous networks.

3 IPv6 Soft handover mechanisms

IP Soft handover approach is based on four main processes, registration process,

duplication process, merging process and handover process. They allow duplication and

merging of IP flows without need to synchronise duplicated flows transmission [11].

3.1 Mobile registration process

In order to be connected to several ARs, MN must be associated with several CoAs, each

CoA identifies MN connection through a unique AR. If we consider a special case of MN

having data connection with two ARs in IPv6 network, and if CN decides to send IP

packets to the MN, sending device have to know all the addresses of MN in all sub

networks. To perform such thing, Mobile IPv6 allows MN to have a primary CoA

(PCoA), which is the temporary address obtained by MN for the first time it connects to

the network. It is registered within home agent and D&M agent in the reference link of

MN and it is the Address used by the different CN, which are likely to communicate with

MN. Two additional local CoAs are used for packets transmission from D&M agents to

MN through the two ARs. Those LCoAs are obtained by MN using IPv6 stateless auto-

configuration addresses mechanism [3] and registered with in D&M agent Figure1

Subnet

PCoA

AR1 AR2

LCoA1 LCOA2

D&M

QoS

broker

Data

exchange

IP network

D&M

Could be

the same

Subnet

Figure 1 Soft handover across subnet.

3.2 Duplication Process

To duplicate packets, D&M agent intercepts all packets sent by the CN and stores them in

its internal memory, extracts from each packet the destination Address (PCoA) and looks

for its corresponding LCoAs. Using those LCoAs, D&M agent creates a new IPv6 packet

with same payload information, but with substitute LCoA as new destination address.

Sequence number (X) is inserted in a Destination Identifier Option (DIO) field and added

to each IPv6 packets header. This field is used to number all packets sent to the tunnel,

same duplicated packets will be identified by same sender, same receiver and same

sequence number. Duplicated and numbered packets are then tunnelled to MN via

corresponding ARs (Figure2). Inversely, MN do same thing with uplink stream. It

duplicates all packets and sends them to the D&M agent via the two ARs.

Access router

Cell A

IP Mobile Node

Cell B

Access

Router B

D&M agent

Correspondent

Duplicated flows

IPv6 flow

PCoA

LCoA (A) LCoA (B)

Figure2: IPv6 flow duplication and merging

3.3 Merging process

The use of D&M agent (respectively MN) duplication process to send separate copies

of same data via multiple ARs to MN (respectively D&M agent), introduces the need to

filter the duplicated packets. To perform efficiently such thing, MN or D&M agent needs

to match those multiples streams in IP layer at reception. In case of uplink traffic, D&M

agent intercepts all tunnelled packets, checks if the DIO field is included in the IP packet.

If there is DIO, which is mean that IP packet was not duplicated, process will route

normally the payload information. D&M agent incorporate a set of tables, particularly a

merging control table (MCT), which defines for each registered LCoA the parameter e

and a list of Xi. e is the highest value X of all received packets plus one. Xi corresponds to

packets that have been transmitted through the tunnel, but which are not yet received.

Those values correspond to packets that are still missing

If DIO is included in the received packet and source-address has an entry in MCT table,

packet has been duplicated, Thus the value of sequence number X and value of e in MCT

table, will be used to determine if this packet is received or not. If received, IP packet will

be discarded (the packet has already been received). Else the payload is routed normally.

Figure3.

3.4 Handover Algorithm

We suppose a MN with two interfaces primary and secondary, the interfaces priority

choice is dynamic; we assume that the primary interface is always the interface with better

connexion quality. The MN must be kept connected through its primary interface. The

secondary interface is used to perform the handover and avoid signal strength degradation

if possible. The aim of this handover strategy is to efficiently exploit all available

resources in order to avoid packet loss and the introduction additional end-to-end delays

during MN roaming from an AR to another one.

Two signal strength thresholds are defined, handover threshold (H_SH), which is the

threshold used in Mobile IPv6 to initiate the handover. Primary threshold (P_SH) is used

in soft handover to initiate secondary interface connection process. Figure 4.

We assume a MN connected on its primary interface with AR1, it has its PCoA and

LCoA1, and both of them are registered with in D&M agent. When MN discovers AR2,

and if quality of primary connexion is less then P_SH, secondary interface connexion is

established with AR2, LCoA2 is registered within D&M, duplication and merging process

will be UP. In this case:

Figure3 : IPv6 flow merging algorithm

Receive packet

DIO

exist…?

Route payload

information to

upper layer

X < e ?

Read value X

yes

Route payload information

to upper layer

Insert intermediate value

Xi in MCT

e = X+1

yes

If X

listed in

MCT ?

Route payload information

to upper layer

Suppress X from MCT

Discard duplicated

packet

Complete process

yes

1. Interface with better connexion quality will be assigned dynamically to be the

primary one. 2. If signal strength of secondary connexion became worst then H_SH, the

secondary connexion is closed and active scanning is initiated to connect it to new AR. 3.

When the Signal strength quality became better then H_SH (very good connexion

quality), MN closes secondary connexion, shut down duplication and merging process.

Complete handover algorithm is described in Figure 5.

MN Primary Interface

Connected to AR1 and

Signal strength > P_SH

Search NAR

To connect

Secondary interface

Connect secondary i nterface

To NAR and start merging and

duplication process

Swappi ng interfaces

Pr i o ri ties

Primary <- secondary

Secondar y<- pri mary

Primary Interface connected

Secondary Interface connected

Disconnect secondary interface

And shut down duplicati on and

merging process

Signal strength

P SH

AR2 s ignal

Yes

Secondar y

interface signal

< H SH

S interface

signal >

P interface

nal

Pr i mary

interface signal

> P_SH

Yes

Ha ndove r

threshold

pr imar y

threshol

Signal

strength

Time

Soft

handover

Ran

Figure 4, 5: Thresholds and Handover process

4 Performance Analysis

4.1 Simulation Environment

To analyze the performances of IPv6 soft handover mechanisms, we use Gemini2

simulator. Gemini2 simulator is a discrete time simulator developed in Eurecom. It

provides support for simple, open and efficient conception of a network topology to

simulate complete wireless networks. Network topology parameters can be chosen at

physicals, data link layers and 802.11 MAC protocol. Above it we have pre-implemented

Mobile IPv6 module over IPv6 routing protocol and UDP is used as transport layer. To

implement soft handover module, we add D&M agent as special router and MN with two

802.11 radio interfaces. No changes have been done to the IPv6 stack. The simulation

model introduces an application in top of CN sending UDP packets to The MN. A number

of IP Access Routers uniformly reparteed give MN optimal radio coverage for about

1000m. A D&M agent is introduced between the CN and ARs in network topology to

duplicate and merge IPv6 flows from CN to MN. A set of MN movement’s scenarios are

used as inputs to the simulation. Each movement scenario determines MN movement at

different speeds across coverage area. The MN1 changes its point of attachment using

basic mobile IPv6 handover and the MN2 is performing soft handover to change its point

of attachment. Figure 6.

Following metrics are used to analyze the performance of soft handover and to

compare it with basic MIPv6 performance: End-to-end transmission delays: the delay

needed by UDP packet sent by CN to correctly reach the application layer in the top of

MN. UDP packets fraction delivery: the number of data packets correctly delivered to MN

over the number of data packets sent by CN. Control/signaling information load: the load

of signaling data generated by MN handover from an AR to new one.

Figure 6 : Simulation network topology.

R9 Ra

Rb Rc

Rd ReR8 Rf

D&M

Virtual

network

2001::E1

Application

UDP

Home

network

2001::E2

2001::E0

802.11 11Mb/s

2001::fa::1

2001::fa::

4.1 Simulation Results

First simulation set aims to determine end-to-end average packets delivery delays

between the CN and the MN. In first simulation set, the MN uses basic mobile IPv6, and

in second set, the MN uses Soft handover mechanisms.

100

150

200

250

mobile speed (m/s)

End to End transfert Dealy (ms)

MIPv6

Soft handover

Figure 5 : Average End to end transmission delays.

By looking at the trends in diagrams 5 showing average end-to-end transmission delays

in both MIPv6 and Soft handover, the following consideration can be made. Soft

handover allows MN to keep a minimal transmission delay, about 25ms when crossing

coverage area. When MN uses MIPv6 basic handover, average transmission delay is to

much bigger, about 170ms transmission. To understand reasons of transmission delays

differences between MIPv6 and soft handover, we plot in diagram 5 and diagram 6 End-

to-end detailed delivery delays for all packets sent by CN to MN, during one MN

movement across the coverage area at 5m/s speed. When MN uses basic MIPv6, the

handover is not initialised before OAR signal strength degradation (it became low than

handover threshold). Before each handover, OAR signal strength degradation generates

successive MAC retransmission of packets before their correct reception. Those

successive retransmissions are responsible of the additional average packet delivery

delays in MIPv6. After each handover, better signal strength from NAR allows correct

reception of packet in Mac layer which avoid additional transmission delay. Delivery

delays of the MN that uses soft handover stays stable, because the MN establishes a

second connection with NAR before severe degradation of OAR signal strength.

Asynchronous emission of duplicated packets through the two ARs allows MN to receive

the first among duplicated received packets at IP layer. That avoids the introduction of

additional end-to- end transmission delays

Figure 6 : MIPv6 End-to-end transmission delays. Figure 7 : Soft handover end-to-end TD

Average UDP packets loss is determined using the same scenarios. The MN moves across

same network topology using soft handover, and after using Mobile IPv6 basic

mechanisms. Figure 7 shows us sum of UDP packets sent by CN and sum of the packets

received by the MN in movement with 5m/s speed. Each handover using MIPv6

introduces packets loss because of 1. Signal degradation before handover 2. MN Layer2

and layer3 disconnection during the handover. The use of two simultaneous connections

in soft handover suppresses packets loss during handover and reduces packets loss

introduced by signal degradation. Several simulation runs with different MN speeds

allows us to have diagram10. By looking at the trends shown in this diagram, the

following consideration can be made.

0 20 40 60 80 100 120 140 160 180 200

100

150

200

250

Time (s)

End to End delivery delay

100

120

140

160

180

200

100

150

200

250

Time (s)

End to End delivery delay (ms)

100

120

x 10

Time (S)

Sum of UDP packets

UDP packet reveive by soft−handover MN

UDP packets sent by CN

0 20 40 60 80 100 120

x 10

Time (S)

Sum of UDP packets

Packet received by MIPv6 MN

packets sent by CN

Figure 8 : UDP received packets in MIPv6. Figure 9. UDP received packets in Soft handover.

By performing Soft handover, MN registers an average of 98% of UDP packets delivery

fraction. This value is stable even MN increase its speed.

When MN uses MIPv6 basic handover, initial delivery fraction is lowest and the increase

of MN decreases the delivery ratio. That decreases from 90% in 5m/s speed to 75% in 16

m/s.

The last simulation set tries to determine and compare load control information

generated by soft handover and MIPv6. To perform such thing, the same simulation

topology is use to evaluate the control load generated by MN handovers. Diagram 11

shows that soft handover introduce additional control load information compares to

mobile IPv6. The additional load is about 40% of basic MIPv6 handovers control load

information.

Figure 10. Average UDP packets fraction delivery Figure 11 : layer 3 Control information load.

Summary and Future Work

In this paper, we have presented a pure Pv6 Soft handover protocol and architecture

that allow MN seamless roaming, with reduced end-to-end transmission delays compare

to various mobile IP approaches. This solution exploit only IPv6 protocols futures,

coexists with MIPv6, improves micro mobility management and can provides data

transmission continuity for delayed constrained applications such as real time video

playback. The Soft handover across heterogeneous networks can be done without any

modifications to MN’s radio system. Comparing to the other IP based soft handover

approaches, the major issues is that there is no need to synchronize the distributed copies

of data. The MN routes first received duplicated packets and simply ignore the others.

We have shown through UDP simulation that IP soft handover is capable, when

enough resources are available, to reduce average packet loss to 2%. It also reduces end-

to-end data transmission delays 6 folds when compared to the standard MIPv6. Those

results show that IP soft handover can be exploited in order to guarantee a high level of

500

1000

1500

2000

2500

number of handover

control information load

MIPv6

Soft handover

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Mobile speed (m/s)

average UDP packets delivery fraction

MIPv6

Soft handover

QoS for real time applications. On the other hand this approach requires the introduction

of D&M agents in network, and introduces additional signal load over the air. Streams

tunnelling and duplication introduces additional overhead of about 48 bytes to each

duplicated IPv6 packets. As future work, we would compare IP soft handover

performance to fast Mobile IP handover (Bi Casting) and smooth handover. Comparison

results will be exploited to develop an adaptive Handover control algorithm across All-

IP networks that guarantee different level of services for applications.

References

[1] Yan Moret & All, European Patent “Process and apparatus for improved communication

between a mobile node and a communication network”, Patent No. 02368057.2.

[2] F.Belghoul, Y.moret, C.Bonnet, “eurecom IPv6 soft handover” ICWN 2003 proceeding, Las

Vegas USA

[3] D. Johnson and C. Perkins. IETF Internet Draft “Mobility Support in IPv6”, draft-ietf-mobileip-

ipv6-24.txt. June 2003

[4] “Wireless LAN Medium Access Control (MAC) and Physical layer”, ANSI/IEEE Std 802.11,

1999 Edition.

[5] Charles Perkins, editor, “IP mobility support for IPv4”, RFC 3220, January 2002.

[6] Charles E. Perkins and Kuang-Yeh Wang. “Optimized Smooth Handoffs in Mobile IP”.

Proceedings of the Fourth IEEE Symposium on Computers & Communications, July 1999

[7] Rajeev Koodli, Charles E. Perkins. IETF Internet Draft "Fast Handovers in Mobile IPv6", draft-

ietf-mobileip-fast-mipv6-07.txt. September 2003.

[8] K.El Malki, IETF Internet Draft ”Simultaneous bindings for mobile IPv6 Fast Handovers ”,

draft-elmalki-mobileip-bicasting-v6-03.txt May 2003.

[9] H. Hartenstein, K.Jonas , M. Liebsch R.Schmitz, M. Stiemerling, D. Westhoff.

“Performance of TCP in the Presence of Mobile IP Handoffs”, (ICT 2001) IEEE International

Conference on Telecom, Bucharest, Romania, June 4-7 2001

[10] Doo Seop Eom, Masashi Sugano, Masayuki Murata, and Hideo Miyahara, “Performance

Improvement by Packet Buffering in Mobile IP Based Networks”, IEICE Transactions on

Communications, vol. E83-B, pp. 2501-2512, November 2000.

[11] Zhang T., Chen JC., Agrawal P., “Distributed soft handoff in All-IP wireless networks”,

Proceedings of IEEE International Conference on Third Generation Wireless and Beyond

(3Gwireless'01), San Francisco, CA, pp. 460-465, mai 2001.