See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/4100782

Improving hard handover in HBR IP micro mobility access networks

employing tree based topologies

Conference Paper · July 2003

DOI: 10.1049/cp:20030364 · Source: IEEE Xplore

CITATIONS READS

2 307

2 authors, including:

Hamed Al-raweshidy
Brunel University London
463 PUBLICATIONS 4,939 CITATIONS

SEE PROFILE

All content following this page was uploaded by Hamed Al-raweshidy on 02 February 2015.

The user has requested enhancement of the downloaded file.

IMPROVING HARD HANDOVER IN HBR IP MICRO MOBILITY ACCESS NETWORKS EMPLOYING
TREE BASED TOPOLOGIES

M. A.O. Barry and H. S. Al-Raweshidy

University of Kent at Canterbury, UK.

Abstract: In IP based wireless access networks, the flow high quality of service, the higher layer end-to-end
of packets to the mobile host (MH) is diverted at the protocols require a certain level of service from
cross-over node (CoN) to the new location of the MH in underlying network protocols to be able to continue
order to reduce handover delay and packet loss. As the providing the level of performance required of them. In
depth of the network increases so does the round trip a fixed wired network, disruption traffic is mainly due
time to the CoN, in the worst case this is the time to the to congestion. As a result protocol stacks at their
gateway or root node of the domain. This paper different layers have been designed and optimised with
proposes an architectural modification to micro mobility this scenario in mind. In a mobile network, however, the
protocols employing virtual or static tree topologies previous premise that disruption in the network is
such as Cellular IP (CIP). We propose an architectural mainly due to congestion is no longer necessarily valid.
extension to the path update process by adding multi- Although for the most part the network is fixed, an
homed base stations (MHOBS) at certain points in the essential stage of the network that facilitates mobility of
wireless access network where the path update delay its users is the wireless stage. It is in this part of the
becomes severe. There are two situations where this network that two primary different causes of disruption
may arise. The first is where the path update is due to a to user traffic appear. The first is the increase in bit error
distant gateway, or other CoN, as a result of an increase rate due to the physical layer of the wireless channel,
in the depth of the topology. The second situation is and hence is dealt with primarily at that layer. The
more likely to arise, with virtual tree topologies on mesh second problem that is faced at the wireless stage of the
networks, when the new path, between new base station network is that of change in point of attachment to the
(BS) and the CoN, is longer than the old path to the network due to mobility. Prior to the Internet and
old BS. The path update minimization proposed here wireless IP-based networks, mobility management was,
will eliminate the effect of path update delay at these by and large, dealt with by layer 2 protocols, with
points in the network, and consequently reduces packet minimal interaction from network-specific layer 3
loss due to the handover process. protocols. This has changed in recent years due to the
increased research in, and the emergence of IP
1. INTRODUCTION mobility protocols.

One of the challenges faced by the future envisaged All- 2.1 Mobile IP Macro Mobility
IP wireless access networks is the smooth handover of
MHs running delay sensitive applications. This is The success of the Internet, that is evident in its
especially true of real-time applications such as ubiquity, has increased the push towards the use of the
streaming and interactive media applications. IP, the layer 3 of the TCP/IP stack, over wireless links.
Handovers are required to be executed with minimal Arguably, this ubiquitous characteristic of IP and its
associated latency and packet loss. Most handovers are potential to provide cheaper packet switched continuous
local by nature, and are confined mostly within the same access to Internet services has motivated the promotion
domain. Therefore the mobility management of the of the all-IP paradigm with the support of standardising
handover can be limited to a local or micro mobility bodies like Internet Engineering Task Force (IETF)[1],
environment as opposed to the global or macro mobility 3rd Generation Partnership Project (3GPP) and
environment. Thus, dividing the mobility management 3GPP2[2]. At the forefront of these protocols is the
problem into two layers of; macro mobility for global IETF’s proposed standard Mobile IP (MIP) [1], Perkins
mobility and micro mobility for regional and local [3]. It is an extension to IP that provides mobility
mobility means that the scope of the handover problem management by using two addresses to maintain
can be narrowed down. services and/or sessions during mobility of the MH in
In the next section IP mobility is reviewed citing the Internet. MIP specifies two main entities called
some examples of the protocols at the forefront of mobility agents; a Home Agent (HA) at the home
research and development in the field. In the subsequent network and a Foreign Agent (FA) at any foreign
sections, our proposal is discussed followed by some network the MH is visiting. When a MH is in its home
results of simulations carried out. network usual IP routing is maintained using its home
IP address, and when MH is roaming it uses the second
2. IP MOBILITY Care-of Address (CoA), which it acquires from an FA.
If none exists the MH can have a co-located CoA. When
The objective of any network is to minimize, as much as a MH migrates to a foreign network, MIP requires that
possible, the disruption to its user traffic. To maintain a the MH must register with its HA after every change to
point of attachment to the network so that packets, visiting MH(s). This effectively means that the mobility
reaching the home network destined for the MH, can be management problem has been divided between two
forwarded to the MH at its new location. This is done levels; macro mobility for global mobility and micro
transparently to the network by encapsulating packets at mobility for regional and local mobility (see Figure 2).
the HA and then decapsulation at the FA or at MH Thus, the micro mobility is managed by specialised
transparently in the case of a co-located CoA. MIP protocols, leaving MIP to manage the macro mobility
requires that the MH register with it’s HA after every problem. Wireless access networks predominantly have
change in its point of attachment to the network. a local or regional scope and therefore mobility
Furthermore MIP mechanism of forwarding packets management in all-IP wireless access networks will be
form any correspondent host (CH) to the MH via the appropriately managed by a micro mobility protocol,
HA causes the creation of the so called inefficient Campbell et al [4].
triangular route between the MH and its CH. (See
Figure 1). HA

Home Domain
CH HA Global
Macro mobility Internet
Home Domain Scope (Mobile IP)
Global
Gateway/ FA
Internet

FA
MH migration Micro mobility
Foreign Domain protocols
scope (CIP)
Handovers
Figure 1- Mobile IP triangular routing. Visited Domain

These two MIP requirement specifications can cause

significant overheads to both the MH traffic and the Figure 2- Macro and Micro Mobility scopes.
load on the core Internet. This is especially true of the
registration requirement as a MH frequently changing 2.2.1 Micro mobility protocols. There are two main
its attachment to the network will be frequently classes of micro mobility protocols that are disparate
registering with its HA. From a network-centric point of based on how traffic forwarding is managed within the
view this frequent registration signalling will increase domain. One class of the protocols uses host specific
the load on the core network. Whereas from a MH routing, also known as Host-Based Routing (HBR), to
traffic point of view registering with a potentially manage traffic forwarding in micro mobility domains
distant HA during frequent network attachment changes Wong et al [5]. Examples include CIP, Campbel et al
will cause continuous disruption to its downlink traffic. [6], and Handoff Aware Wireless Access Internet
Eventually this will lead to loss of upper layer (transport Infrastructure (HAWAII) Ramjee et al [7]. The other
and/or application) connectivity due to this poor quality class uses Tunnelling as in base MIP. Examples are
of service from the IP layer. Hierarchical MIP (HMIP) Soliman et al [8], and
TeleMIP/Intra-Domain Mobility management Protocol
2.2 IP Micro Mobility (IDMP) Das et al [9]. Although our delay minimization
can be applied to both of the two classes of protocols, as
The delays to traffic destined to the MH due to MIP’s it is an architectural optimisation, an HBR protocol
requirement that it perform a registration after every based on CIP was used in this investigation. This is
change in its network point of attachment, and the high because of the inherent advantage in HBR schemes
latency due to triangular routing to a potentially distant manifest in the integration of routing and location
HA, make MIP inherently less suited to frequent management, Campbell et al [4] Wong et al [5]. This is
changes in the network point of attachment of MHs in contrast to the two-tire approaches of tunnelling
within a domain, i.e. intra-domain. This type of intra- protocols that are derivatives of MIP. In a domain using
network mobility has been termed micro mobility a HBR protocol, soft-state routing functionality is used
Campbell et al [4]. A number of micro mobility coupled with that of a cross-over node (CoN) in the
protocols have been proposed. These protocols rely on mobility management of problems of location updates
the notion that frequent mobility by its nature is a local and traffic redirection, as a result of frequent MH
or regional phenomenon. As a consequence of this handovers. A CoN is the node that lies on the closest
premise the inherent triangular routing of the base MIP intersection between the old and new paths to the MH.
rarely becomes an issue. More crucially, frequent While Soft-state routes are those of MH specific routing
changes to the MH network point of attachment no information that include MH address. These soft state
longer mean frequent MIP registrations. This is because routes are created dynamically on a hop by hop basis
MH mobility is now managed within the domain by the with uplink control packets, called route or path
micro mobility protocol, transparently to MIP, with the updates, in an explicit manner. They are subsequently
gateway or root node of the domain acting as the FA for explicitly refreshed by the path updates as well as
implicitly refreshed by uplink data packets from the BS. This could be the case in an inter-subnet handover
MH. If the soft-sate cache entry is not refreshed by a when the Gateway in acting as a CoN, as illustrated in
certain route-update time they expire and are erased Figure 3.
from the routing state in the respective node. In this Gateway
New path
manner HBR protocols enable the access network to (New CoN)
(Old CoN)
combine routing with location management for active Routers Old path
MHs, while at the same time the soft-state aspect makes Hierarchy
the usually unacceptable scaling disadvantage of host- levels

specific routes in the global Internet, an advantage in the

local scope of the micro mobility domain, BS 5
BSs BS 4
Wong et al [9].
Handovers
2.2.2 Micro mobility management. Mobility
management includes three main aspects, Eardley et al Figure 3- Inter-subnet handover in the access network.
[10]. The first two are concerned with the management
of active MHs, while the third deals with the The arrows pointing up represent uplink path update
management of Idle MHs that are connected to the packet that the MH transmits after every handover, with
network but not engaged in sending and receiving user the arrows pointing down represent downlink traffic
data. These aspects are: flow to the MH. During handover, the packet loss will
 Path Updates. be proportional to the round trip time to the gateway,
 Paging and the downlink packet rate. Our proposal is to modify
 Handover Management. the topology by adding Multi-homed BSs (MHOBS) at
The Path Updates aspect is absolutely required, as it these and other depths to counter act the effects of
is how the network knows the location of the active relatively distant CoNs on the handover process.
MH, and hence knows where to route its traffic to. Our MHOBS addition will ensure the elimination of packet
proposal relates to handover which is an active host loss, during transmission of the location update packet
issue. This is opposed to paging which is concerned to the gateway, by the elimination of the effect of the
with idle hosts. Thus the paging process is outside the route update delay on the handover process altogether at
scope of this investigation and will not be that depth in the hierarchy. So the handover associated
mentioned further. with the maximum path update delay, as shown in
Handover management has two aspects; an Figure 3, is replaced with 2 handovers with the
architectural or network-based aspect, and a protocol or minimum delay, as illustrated in Figure 4.
signalling aspect. The network-based aspect relates to
Gateway
whether a particular topology is required, for example
static/virtual tree requiring CoN functionality. The (Old CoN) (New CoN)

second aspect, the algorithm signalling aspect, relates to MHOBS

the actual handover scheme’s control packet exchanges

in executing the handover signalling. For example,
whether buffering and redirection of packets is required.
As the scope of our proposal is architectural, it is
directly related to CoN functionality, and as such, the
Handovers
second signalling aspect in this investigation is limited
to a validational role. Figure 4- Inter-subnet handover using MHOBS.

3. PATH UPDATE MANAGEMENT EXTENTION 3.1 Offsetting path update delay in handover

In a HBR domain, to reduce handover delay and packet CIP semi-soft handover scheme has an option to bi-cast
loss, the flow of packets to the MH is diverted at the path updates from a MH to two BSs to facilitate fast
CoN, which is nearest node at which the old and new handover Campbell et al [4]. We propose taking the bi-
paths intersect, and in the worst case this is the gateway. casting functionality up one level from the MH, to the
Deploying the network over as wide an area as possible BS regardless of the handover scheme whether hard or
is desirable as it keeps the mobile user in the same semi-soft. A larger illustration showing the paths
network for a longer amount of time. As the depth of the between the CoNs and the MHOBS is shown in
network increases so would the round trip time to the Figure 4.
CoN, depending upon its level in the hierarchy. This Since this is an architectural modification it can be
increase in the depth of the topology is significant since implemented in any micro mobility protocol employing
control packets that traverse the network form the new a tree based topology and not only limited to CIP.
BS will take longer to get to the gateway to modify the Unlike in CIP where bi-casting is implemented at the
MH cache entry and create a new path pointing to the MH at every handover, in our proposal bi-casting is at
new location of the MH. All the while packets already specifically placed MHOBSs to minimize route update
transmitted are being lost through the old path at the old times in due to any increase in the depth of the
topology. Reducing the time required to update paths to 4.1.1To-and-fro scenario simulations. Two models
the new location of the MH means the handover can be were simulated for comparison. One to show the
executed faster i.e. in less time. Hence, having the MH performance of a topology with MHOBS, and the other
routing state information as close as possible to the MH topology without MHOBS. The topology with MHOBS
location will facilitate a faster handover since the route is shown in Figure 5, while the one without is shown in
update can reach the CoN sooner. At its closest, the Figure 6. The two models were identical in terms of
CoN will be the parent of both old and new BSs. This is network parameters. A session of 80s was simulated
the configuration that yields the minimum path update with 12 handovers every 5s, starting from 15s to 75s. In
time, hence the MHOBS is able to offset the latency of the simulation without MHOBS the handovers were to-
the path update time to this minimum. The MH will and-fro between BS2 and BS3. While in the topology
continue to receive traffic from the old CoN while the with MHOBS the handovers were from BS1 to BS2,
new one is being configured along with the new routes BS2 to BS3 and back the same way to BS1. To and fro
during the path update process. Figure 3 shows a typical handover scenario method was used so as to bring out
tree topology of a micro mobility wireless access the performance improvements of MHOBS.
network domain. The use of MHOBS is not only limited
(New CoN)
to when the gateway is the CoN. Indeed, as the depth of Old path New path
the tree topology increases, the MHOBS can be used at
(Old CoN) Gateway
other levels to offset the route update delays to
respective CoNs. Figure 4 shows the modified topology
employing a MHOBS to offset the path update delay to
the gateway and back.
BS1 BS2 BS3 BS4
Gateway
New path
(Old CoN)
Old path (New CoN)
Figure 6- Handover without MHOBS topology used in
the simulation.

BS3 Figure 7 presents the number of packets lost in total

BS1 BS2
after the eight simulations (run with varying values for
the link latency for each simulation). It can be seen that
Figure 5- Optimised Handover with MHOBS connected the MHOBS enabled topology resulted in loss of fewer
to both the old and new CoNs. packets in almost all the cases as compared to the
topology without MHOBS. On average more than 100
Where there would usually be two BSs the MHOBS is additional packets were lost in simulations without
used to eliminate effect of the handover process when MHOBS. The worst case packet loss for MHOBS was
the gateway is the CoN. Therefore the increase in the 150 packets compare with almost 350 without MHOBS.
depth the topology, and as a consequence the increase Indeed, the worst case packet loss for MHOBS was only
distance of the CoN (in this case the gateway), is slightly more than the best packet loss for the one
effectively reduced to the minimum. Thus the handover without MHOBS. For the link latencies of 2ms about
is reduced from a level-3 CoN (the Gateway) shown in 140 packets were lost compared to the MHOBS packet
Figure 3, to one with a level-1 CoN shown in Figure 4. loss of only 3 packets for the same link latencies.
Figure 5 shows a close up view of the signalling and
traffic flows.

4. PERFORMANCE EVALUATION

4.1 Simulation Environment

The model used for the simulations assumes cells with

perfect overlapping coverage, zero transmission errors
and zero propagation delay. Lower layer 2 handover is
assumed to be smooth and instantaneous. Other
parameters used to facilitate the simulations are
bandwidth of wired and wireless links, latency of wired
and wireless links, and average size of path update and
data packets. Traffic was constant bit rate UDP to
simulate voice application with a data rate of 25kbits/s.
The model was built and simulated using OPNET
Modeler [11]. Two different scenario sets are Figure 7- Packet losses for 8 simulations after 12 to-
presented next. and-fro handovers.
 4.1.2 Realistic simulation scenario. Another set of 4. A. T. Campbell, J. Gomez, S. Kim, Z. Tur´anyi, A. G.
simulations were run for a more realist movement Valk´o, C-Y. Wan, 2002, “Internet micromobility”,
scenario. This time the movement was not restricted, Journal of High Speed Networks, IOS Press, (11).
and the handover were between all the BS in the
respective topologies. These were similar to the ones 5. K. D. Wong, H. Wei, A. Dutta, K. Young, 2001,
shown in Figures 3 and 4 respectively, except that for “Performance of IP Micro-Mobility Management
the MHOBS scenario, there were 3 MHOBS and 4 BS. Schemes using Host Based Routing”, Proceedings 4th
From Figure 8, it can be seen that the MHOBS in these International Symposium on Wireless Personal
more realistic simulations still perform better than the Multimedia Communications (WPMC'01).
topologies without MHOBS. For example after 4
handovers of a 40s session, the 100ms link latency 6. A. T. Campbell, J. Gomez, S. Kim, A. G. Valk´o, C-
MHOBS dropped 155 packets compared to 175 packets Y. Wan, and Z. Tur´anyi ,2000, “Design,
in the simulation without MHOBS. While the Implementation, and Evaluation of Cellular IP”. IEEE
improvement in reduction in packet lost is not as Personal Communications,7(4):42–49.
dramatic in these sets of simulations as compared to the
to and fro sets, it still is a significant improvement for a 7. R. Ramjee, T. La Porta, S. Thuel, K. Varadhan, L.
more realistic scenario. Salgarelli, 2001, “IP Micro-mobility support using
HAWAII”, IETF draft. (work in progress)
http://www.ietf.org/internet-drafts/draft-ietf-mobileip-
hawaii-01.txt

8. H. Soliman, C. Castelluccia, K. El-Malki, L. Bellier,

2001, “Hierarchical MIPv6 mobility management
(HMIPv6)”, Internet Draft (work in progress)
http://www.ietf.org/internet-drafts/draft-ietfmobileip-
hmipv6-04.txt

9. S. Das et al., 2000, “TeleMIP: Telecommunication-

Enhanced Mobile IP Architecture for Fast Intradomain
Mobility”. IEEE Personal Communications, 7(4):50–58.

10. P. Eardley, N. Georganopoulos, M. West , 2002

“On the Scalability of IP micro mobility management
protocols”, IEEE Conference on Mobile and Wireless
Figure 8- Packet losses for 6 simulations after 4
Communication Networks (MCWN2002).
handovers in the realistic scenarios
11. http://www.opnet.com
5. CONCLUSION
In this paper, an overview was presented of the IP
mobility paradigm, highlighting the difference between
the macro and micro mobility approaches. Then a novel
path delay minimisation was introduced, discussed and
simulation results have been presented. It has been
shown that the proposed MHOBS-based approach
enabled topologies had less packet loss during the
handovers. This was more pronounced in the to-and-fro
scenario when compared to other configurations. The
simulations clearly show the extent of the
improved performance.

REFERENCES

1. http://www.ietf.org/

2. http://www.3gpp.org/

3. C. Perkins, 2002, IP Mobility Support for IPv4 , IETF

RFC 3220, http://www.ietf.org/rfc/rfc3220.txt

View publication stats