1. Introduction
In cellular networks such as GSM, a voice call is seamlessly handed over from cell
to cell using hard handover1. This is possible because the cells are part of the same network
and network-based handover control mechanisms detect when a user is in a handover zone
between cells, by measuring the received signal strength at the radio interface in the
Mobile Node (MN). These network-based handover control mechanisms use this
information to determine when to redirect traffic to the cell that the MN is moving into.
As cellular networks evolve to enable IP-based traffic through the General Packet
Radio Service (GPRS), for example, and with the advent of 3G networks, the management
of handovers between cells becomes somewhat more difficult. When a user moves
between cells in a GPRS enabled network, there will often be packet losses and an
associated interruption in multimedia delivery. This is addressed in 3G networks by
allowing the MN to attach to two base stations simultaneously during the handover period.
This is known as soft-handover and requires the MN to manage the timing of the handover
between the two cells.
It is expected that the widespread deployment of Wireless LANs (WLANs) and
other Radio Access Networks (RANs) and possible cellular evolutions beyond 3G will lead
to a heterogeneous wireless network environment, often referred to as 4G [1]. This will
require the handover of sessions between different networks rather than simply between
cells in the same network. These handovers are usually referred to as vertical handovers.
The work presented in this paper examines the issues involved in performing a soft
handover of a video streaming application in a heterogeneous wireless network
environment.
The paper is laid out as follows. The next section presents some possible handover
strategies, while the third section presents the rationale for using delay as a metric for
deciding when to handover. The proposed scheme is then described in detail and results are
presented from a test bed that emulates network conditions. Some preliminary delay
1
A handover in which there is only one connection carrying data at any one time.
measurements in a live WLAN are then presented. The final section presents the
conclusion and outlines our plans to develop this work.
2. Handover Strategies
There are a number of proposed methods for performing a handover in a wireless
environment operating at different layers in the various protocol stacks. Most are designed
for a single access radio technology and hence are not suitable in a heterogeneous
environment. Approaches that operate above layer two are most suitable for handover in a
heterogeneous environment. Here we discuss the main proposals for handover above layer
two.
An approach which has received most interest within the research community is
that of Mobile IP [2], which enables a MN to receive IP packets through a packet
forwarding procedure. This approach is well suited to solving the problem of locating a
mobile that may be attached to one of a number of networks. However, handovers in
Mobile IP are slow and packets can be lost during the handover procedure [3], making it
unsuitable for handover of video traffic.
Extensions to the mobility support provided in the Session Initiation Protocol (SIP)
[4] have also been proposed for delay-sensitive communication [5,6]. SIP is an application
layer protocol for establishing and tearing down multimedia sessions. It can help provide
personal mobility, terminal mobility, and session mobility. SIP uses hard handover and
“in-flight” packets can be lost during the handover period. In [5] it was estimated that
packet loss can occur for up to 50ms during the handover period. This estimate did not take
into account network initialisation procedures such as AAA or address assignment, and
this can be a major part of the overall handover delay. SIP therefore is not good at
providing seamless handover of streamed video in networks.
Clearly, the above solutions are inadequate for handover of streamed video,
primarily because they break the old connection before they make the connection to the
new cell. In soft handover the MN is connected simultaneously to two Access Points2
(APs) and the same data packets are sent through each of the APs during the handover
period. The MN can use either stream and when a stream has not been used for a period of
time, the MN drops that connection. This means that packets are not lost during the
handover and there is no interruption to service, making it suitable for handover of video
traffic. The price to be paid for seamless video service is the use of resources in two
networks, rather than one, but only during the handover phase.
An approach to realizing handover based on a new socket abstraction was proposed
in [7]. While this is an interesting contribution it did not address when to handover
satisfactorily. In this paper we use a delay-based scheme that enables a MN to make this
decision for the handover of a multimedia stream between two wireless networks. This
approach makes the strategy independent of the radio technology used and does not require
cross-layer signalling.
2
The term is used here to indicate the point that provides access to the network infrastructure, e.g. a Base
Station in GSM, an Access Point in Wireless LANs, etc.
measurement-based technique is used to assess which network has the better performance
and then use the corresponding stream.
When the distance between the AP of the wireless network and the MN is quite
large, the signal weakens to a point where packets need to be retransmitted by the MAC
layers. If the video is transported over TCP, further retransmissions at the transport layer
occur if the retransmissions at the MAC layer fail. As this distance increases, the number
of retransmissions required generally increases. The net effect is a reduction in effective
throughput and an increase in packet delay [8]. Similarly, a weakening of the signal due to
obstacles between the AP and the MN can result in retransmissions.
It is interesting to note that in wireless packet switched networks, such as WLANs
and GPRS, distance from the AP is not always a good indicator of network performance. A
MN located close to the AP in a WLAN cell that is busy with traffic from other MNs in the
same cell may receive a small share of the total bandwidth. On the other hand, a client
located further away from the AP in a less busy WLAN cell may receive a greater amount
of bandwidth and its packets can therefore experience less delay.
Packet delay is a metric of network performance and can be used in a soft handover
procedure to decide when to handover to another wireless network. In [9] a modified
transport layer protocol used path delay to initiate a handover with a voice application. The
work presented here investigates handover of a video stream in the application layer, using
a delay metric to initiate the handover.
The MN applies hysteresis to the delay difference, by having a threshold that the
delay difference must exceed before a handover can proceed. This prevents ping-ponging
between the two networks when the congestion levels are similar in both networks.
An application layer agent was created to implement this scheme and was used for
gathering the results presented below.
5. Emulation Results
To validate the concept of this approach, the proposed scheme was tested in the test
bed illustrated in figure 2. The test bed uses Nistnet [10], which is a network emulation
tool that is frequently used for emulating network conditions at the IP layer. It can emulate
network conditions (e.g. packet delay) on a per path basis. The server and MN were
connected to the computer running Nistnet using Ethernet links. Two separate paths were
used to emulate two RANs and the MN had two IP addresses.
Ethernet Ethernet
Server Nistnet MN
The server transmitted two identical streams of Real Time Protocol (RTP) packets
[11] at 90kbps using the Transport Control Protocol (TCP). The packets pass through
Nistnet on their way to the MN and the MN sends no packets to the server.
Nistnet was used to emulate packet delay that might be expected in a RAN that
experienced temporary congestion. As figure 3(a) shows, Nistnet applied a constant delay
of 40ms to all packets arriving from Stream 2. For the first 108 packets in Stream 1 it
applied no delay. It then applied a constant delay of 80mS to all packets in Stream 1 up to
packet 196. For the remaining packets from Stream 1, no delay was applied.
The diagram in figure 3(b) shows the delay difference generated by the MN and the
handover decision made by the MN based on this delay difference. As figure 3(c) shows,
the client only performs a handover when the delay difference exceeds the hysteresis
thresholds.
The above experiment was repeated a number of times, with consistent results.
The choice of threshold levels are issues for further study.
thresholds
Stream 2
Stream 1
5. WLAN Results
In order to evaluate the use of delay as the basis of a handover mechanism for a
roaming MN in an uncongested WLAN, experiments were performed to investigate the
relationship between distance from the AP and delay in an uncongested WLAN. Both line-
of-sight and non-line-of-sight scenarios were examined.
A video server and MN were connected to an 802.11b WLAN as shown in figure 4.
The server streamed RTP packets over TCP to the MN via the AP. This RTP stream was
used to model CBR video traffic. The packets were transmitted at 30 packets per second
and contained 1448 bytes of RTP data each, corresponding to a video bitrate of 347,520
bits/s. The MN was initially beside the AP, and was then moved at a steady walking pace
away from the AP until the range of the AP was exceeded.
Video WLAN MN
Server
Wire
Wireless
Figure 5 shows a graph of the delay experienced by the RTP packets when the AP
and MN were located outdoors with line-of-sight between the AP and MN throughout the
test. The X-axis roughly corresponds to distance between the MN and the AP, where the
first packet on the X-axis corresponds to when the MN was beside the AP, and the last
packet corresponds to when the MN was about to go outside the range of the AP. It shows
that the packets experienced no delay until the MN was approaching the edge of the
WLAN cell, and that the delay increased exponentially as the MN reached the edge of the
cell.
Figure 6 shows a graph of the delay experienced by the RTP packets when the AP
and MN were located within a building and there was no line of sight between the AP and
the MN. Again, the X-axis roughly corresponds to distance between the MN and the AP.
Analysis of the data in the tests showed that each instance of high delay was caused
by TCP retransmissions. The results above suggest that delay is suitable in deciding which
is the better network during a soft handover.
6. Conclusions and Future Work
In this paper we explained why soft handover is appropriate for the handover of
video between networks in a heterogeneous wireless network environment. We showed
how relative packet delay between the two paths can be used by the MN to determine
network performance and how the MN can use this during a soft handover to determine
which is the better network. We described a scheme that can be implemented in the MN to
achieve this. Finally, we presented results showing successful handovers in an emulated
environment and delay measurements in a live WLAN.
We are currently working on using the scheme to demonstrate handovers between
WLAN cells. Later, the use of the scheme to handover between a WLAN and other
wireless networks, such as GPRS, will be examined.
Acknowledgements
The support of the Informatics Research initiative of Enterprise Ireland is
gratefully acknowledged. Thanks to Dr. Sean Murphy for reviewing the paper.
References
[1] Hui & Yeung, “Challenges in the Migration to 4G Mobile Systems”, IEEE Comm.
Mag., Dec 2003
[2] C. Perkins, “IP mobility support”, RFC (Proposed Standard) 2002, IETF, Oct 1996
[3] Stephane & Aghvami, “Fast Handover Schemes for Future Wireless IP Networks: a
Proposal and Analysis” , VTC, May 2001
[4] M. Handley et al, “SIP: session initiation protocol”, RFC 2543, IETF, Mar 1999
[5] Wedlund & Schulzrinne, “Mobility Support using SIP”, Proc. ACM WoWMoM'99,
USA, Aug 1999
[6] Wedlund & Schulzrinne, “Application Layer Mobility Using SIP”, Mobile Computing
and Communications Review, Volume 1, Number 2, 2001
[7] J. Kristiansson and P. Parnes, “Application-layer mobility support for streaming real-
time media”, Proc. IEEE Wireless Communications and Networking Conference,
2004
[8] Liu et al, “Towards High Performance Modeling Of The 802.11 Wireless Protocol”,
Proceedings of the Winter Simulation Conference, December 2001
[9] Noonan et al, “Simulations of Multimedia Traffic over SCTP modified for Delay-
centric Handover”
[10] Nistnet emulation tool, http://snad.ncsl.nist.gov/itg/nistnet/
[11] Schulzrinne et al, “RTP: A Transport Protocol for Real-TimeApplications”, IETF
RFC 1889, January 1996