Path Location Register for Next-Generation

Heterogeneous Mobile Networks

Theodore Zahariadis

, Stamatis Voliotis

Ellemedia Technologies/Bell Labs, 223 Syggrou Av.,Athens, GR-171 21 Greece

Technical Educational Institute of Chalkida, Psahna, Greece

Abstract. Deployment of a global all IP wireless/mobile network is not a

straightforward decision. Heterogeneous mobile networks combined with

wireless “hot-spot” locations seams to be one of the most realistic early

deployments. Commercial public wireless LAN solutions however offer

proprietary location management capabilities compared to the traditional

cellular networks. The increasing demands for heterogeneous services

necessitate fast and efficient location management mechanisms that allow the

future personal communication service network to locate mobile users roaming

across different systems. This paper introduces and analyzes a Path Location

the cost of mobile terminal location update and paging. The performance

evaluation of the PLR scheme demonstrates its effectiveness in next generation

heterogeneous mobile networks.

1. Introduction

Forth generation (4G) all-IP networks are expected to provide a substantially wider

and enhanced range of interactive multimedia services. Terminal and personal

mobility will enable users to access their personal profile, independently of the

terminal type or the point of attachment to the network. However, deployment of a

global all-IP wireless/mobile network is not a straightforward decision, due to

technical and economical issues. A phased approach, integrating heterogeneous

2G+/3G and wireless LAN technologies on “hot-spot” locations, appears to be one of

the most realistic early deployment approaches. In order to facilitate global

connectivity with maximum bandwidth and minimum cost a variety of mature

wireless/ mobile technologies can be considered. In the local area, the Wireless LAN

(WLAN) is a well-established and expanding market, with superior bandwidth

compared to any cellular technology and supported by international standards (i.e.

IEEE 802.11 a, b, g, e, ETSI HiperLAN I & II, Bluetooth). Regarding the wide area

network, mature cellular standards are already deployed (i.e. GPRS, EDGE, IS-95,

CDMA). In case of absence of cellular network, satellite links can fulfill the

requirement for worldwide coverage [1].

Connectivity at the physical layer is mandatory, but this is only a part of the

problem. The increasing demand for heterogeneous services necessitates fast and

efficient location management mechanisms that allow the future personal

Zahariadis T. and Voliotis S. (2004).

Path Location Register for Next-Generation Heterogeneous Mobile Networks.

In Proceedings of the 3rd International Workshop on Wireless Information Systems, pages 142-151

DOI: 10.5220/0002670301420151

 SciTePress

communication service (PCS) network to locate mobile users roaming across different

systems. Generally a location management scheme contains two processes: location

update and paging. In case of uniform systems, many location management schemes

have been proposed and evaluated for both cellular systems [2][3] and computer

oriented networks [4][5]. In case of heterogeneous PCS systems, the registration, call

delivery and handset identity are discussed in [8], while methods for enhancing the

network’s location management in multitier (GSM, IS-95, IS-54) systems have been

proposed in [6][7]. Roaming across systems imposes a significant increase in

signaling traffic. However, 3G+ and 4G Mobile Networks will not be voice-centric,

but QoS aware data centric; thus specific location management algorithms that take

into account parameters like QoS, call and packets loss, paging delay should be

considered. In this paper, a Location Management scheme for heterogeneous

networks is analyzed and evaluated. The scheme is based on the introduction of a

layer of Path Location Register (PLR) servers along with roaming and paging

algorithms that handle mobile terminals mobility on local or regional base. The

performance evaluation of the proposed scheme demonstrates its effectiveness in

heterogeneous next generation mobile networks.

2. All-IP heterogeneous network architecture

In a multitier system consisting of heterogeneous wireless technologies, different

networks are combined in order to cover a specific geographical area. Each network

may comply with different specifications and standards, and encompass different

number of cells, while cell overlapping is expected. Cells’ physical or logical

diameters and transmission characteristics (e.g. bandwidth, maximum number of

terminals, connection set-up time, call tear-down probability) may vary. Even in the

same tier, parameters like the signaling messages sequence and format, the

authorization rights etc. may differ.

For example in Fig. 1, three different networks are shown. A mobile terminal

(MT) may roam between cells of the same tier or between cells of different networks.

In case MT enters an area where cells overlap, it may select to handoff to the newly

entered network or remain attached to the previous one. The handover selection may

be based on the networks’/ cells’ characteristics or on the network handoff/location

management overheads. As shown, an extended Home Location Register (HLR+)

may control a number of different networks either of the same or of different types.

When a terminal roams between cells of the same type, it may or may not change

servicing area (and Visiting Location Register, VLR). For example, when the MT

roams from cell A to cell B, it does not change servicing area or network. When it

roams from cell B to C, it changes cell, servicing area and network; thus it changes

from VLR2 to VLR3. Moreover, when the terminal roams from cell D to E, it

changes VRL and HLR+, though it does not roam to a new wireless network

technology.

143

Network A

Network C

VLR 3

HLR+ 1HLR+ 1 HLR+ 2HLR+ 2

VLR 1

VLR 2

VLR 6

VLR 5

VLR 4

Network B

Fig. 1. Network Hierarchy Architecture

The problem in the above architecture is that the respective VLRs and HLR+ have

to be updated every time the MT roams to a new cell. This heavily increases the

signaling overhead especially in case the MT moves back and forth in the

surroundings of a servicing area, the so-called “ping-pong” effect. Apart from the

extensive signaling, the ping-pong roaming effect causes additional overheads, due to

the locality of the IPv4 addresses. Mobile IP and various alternatives and extensions

[8] aim to face the problem of mobility in both wireless and mobile environment, but

none has yet managed to take into account the mobility management, the QoS

requirements and the heterogeneity of the network, while other (i.e. HAWAII [9],

Cellular IP, UniWA [10]) use layer-3 signaling, increasing the handover latency and

originating significant packet losses.

3. Path Location Register

Aiming to solve efficiently the mobility problem in heterogeneous all-IP

wireless/mobile networks, we introduce a Path Location Register (PLR) management

scheme. The scheme includes an intersystem roaming and a paging algorithm.

In the proposed schema, a PLR servers’ layer is introduced in lower hierarchy from

the VLR that trail the MT when roaming, primarily on the boundaries between IP

networks. Terminals are assumed to be multi-band/multi-standard devices able to gain

connectivity either in macrocell or microcell environment. Each MT is permanent

associated with an extended HLR+. When the MT moves to a visiting network, it is

temporary assigned to a VLR, which updates the HLR+ for the terminal position. In

parallel, a PLR is also informed in order to keep track of the MT movements in local

basis. When the MT roams to a neighboring cell, the PLR may continue to route

traffic to the terminal either directly or via a PLR that is “closer”. The distance

between the terminal and the PLR may be defined as a function of the cell

characteristics (diameter, load, current number of terminals, QoS capabilities), the

terminal motion (speed, direction) or the call requirements (bandwidth, handoff

sensitivity, error correction). As the traffic is routed via the PLR, it can easily track

the MT and inform the neighboring PLR when the MT is approaching the servicing

area boundaries. When the MT roams to a cell or network of different type, the PLR

may handle additional issues, such as air interface compatibility, user/terminal

authentication, billing etc.

144

Network A

Network C

VLR 3

HLR+ 1HLR+ 1 HLR+ 2HLR+ 2

VLR 1

VLR 2

VLR 6

VLR 5

VLR 4

Network B

PLR-2

PLR-3

PLR-4

PLR-5

PLR-6

PLR-8

PLR-1

PLR-7

Fig. 2. Path Location Register Network Architecture

For example in Fig. 2, when the MT is located in cell A, it is also assigned to

HLR+ 1, VLR1 and PLR3. The same PLR keeps tracking the MT and routes traffic

until it reaches position C. As the networks A and B overlap, the MT may decide to

avoid roaming to Network B, but continue to communicate via PLR3. Nevertheless

the PLR4 is informed that the terminal has entered its servicing area, so according to

MT move, call requirements and network load, the PLR4 performs a preliminary

resource allocation in the neighboring cells. For instance if Network B is a public

Wireless LAN, the MT may select to keep the cellular interface active, while in

parallel the MT’s 802.11 interface and the network access node are prepared for a

potential handoff. In this way, if the MT returns to cell B no actual roaming is

performed, while if the MT roams to cell D, traffic is routed via PLR4. As PLR3 and

PLR4 belong to the same HLR+, they are considered “close by”; thus the VLR layer

is not informed at all, while the roaming is handled in PLR layer. According to

network ownership, these PLRs may be considered “close by” or “remote”. For

example, if the MT roams to cell E, traffic may be routed via direct links between

PLR3, PLR4 and PLR5, or the VLR and HLR+ hierarchy may be informed. The

drawback is that inter-PLR links increase the paging delay; thus thresholds in PLR

links paths are introduced.

145

Start

Yes

Calculate

PLR distance

Create a new

PLR path

PLR distance < Dp

Locate an

appropriate PLR

Yes

Inter-PLR path

length < D

Creat e a new

Inter-PLR link

Create a new

PLR path

Update the VLR &

the HLR servers

Start

Yes

Is the called terminal

at the same PLR?

Is the called terminal

connected via an

inter-PLR link?

Yes

Follow the path

to the PLR

Page/Traffic

Route Directly

Follow the inter-

PLR link(s)

Yes

Is the called terminal

at the same

VLR/HLR?

Follow the path

to the VLR/PLR

a) PLR Roaming Algorithm b) PLR Paging Algorithm

Fig. 3. PLR Roaming Algorithm & PLR Paging Algorithm

In Fig. 3a the PLR roaming algorithm flow diagram is shown. For simplicity we

assume that there are no ownership, deployment or other issues, but the decision for

handoff is based only on performance criteria. As shown two thresholds are

measured: the D

, which is the maximum allowed distance between the MT and the

PLR and the D

IP,

which is the maximum length of the inter-PLR link. In classical

roaming algorithms, the VLR and the HLR+ should know the path to the terminal in

order to be able to route incoming calls and packets. This would result in many

routing entries updates, and many VLR and HLR+ signaling messages. In order to

minimize this overhead in the PLR scheme, we postpone the HLR+ update and treat

the roaming in local or regional layer. When a terminal roams to a new servicing area

the distance between the servicing PLR is checked. If it is less than the maximum

distance D

a new PLR route is created and no further actions take place. If the

maximum distance is exceeded, an “appropriate” new PLR server is located and an

inter-PLR link is created. Many criteria can be involved in the selection of the new

PLR: the distance from the terminal, the location, the terminal’s call and connection

requirements, the PLR load, or even statistical measurements and profiles may be

involved. For simplicity reasons each PLR has a list of neighboring PLR’s, so

searching is efficient. When creating the inter-PLR link, an optimal routing algorithm

may be invoked to check if the path has some cycles, or if a path between the PLR

servers already exists. If no “appropriate” PLR can be found, the VLR and HLR+ are

informed and the complete path to the terminal is refreshed.

146

Due to PLR roaming algorithm, the paging algorithm is also modified.

Additionally to direct indexing from HLR+ to MT, we have to trace the PLR and the

inter-PLR links if exist. As shown in Fig. 3b, the paging/traffic routing algorithm

starts from the PLR that the MT is located, and follows a bottom up approach. If the

intermediate layers fail to locate the MT in the servicing area, the HLR+/VLR layer is

reached and normal routing is followed. The paging delay is the overhead, the PLR

has to pay for benefit of less signaling at the roaming phase. However, if the paging is

not a critical factor, the longer the inter-PLR link chain, the largest saving could be

obtained. In some cases due to the heterogeneity of the network, the PLR

paging/traffic routing algorithm may be even more efficient than normal paging, as it

assumes larger servicing areas and omits searching in adjacent network systems.

The main benefit of the PLR scheme is that it significantly reduces the signaling

cost and the set-up overhead caused by the intersystem roaming. Traffic routing is not

modified, but the network traces the MT as it moves from cell-to-cell and from

network-to-network and adds or drops links and paths accordingly. Moreover,

roaming is handled locally in each servicing area, so the ping-pong effect is omitted.

Another advantage of the PLR scheme is that the additional layer of PLR servers does

not affect the original database architecture. The additional hardware and

communication links between PLRs can be safely balanced by reducing the number

of VLR servers in an area.

4. Performance Analysis

In this section we adapt the analytical model of [12] in order to evaluate the

performance of the proposed PLR scheme. Lets assume that the calls towards a

terminal have mean rate λ and the mean time a terminal is located in the servicing

area of a PLR is 1/µ. Then the terminal call-mobility ratio (CMR) in this area would

be CMR=p=λ/µ. If the PLR algorithm is not applied, the HLR+ and the VLR servers

will be informed every time the terminal roams to a new cell. Otherwise it will be

informed each time the path to the terminal exceeds a maximum distance of D

PLR

+ D

, where D

is the length of inter-PLR links and D

is the distance between

the last PLR and the terminal. If we assume that by average the terminal changes PLR

every T

moves and the D

has a length of T

links, the D

PLR

distance will by

average result after T

moves assuming that no circles are measured.

If the user roams to n different PLR servers between two calls the HLR+ will be

updated N

HLR

⎥

⎦

⎥

⎢

⎣

⎢

PIP TT

times. The number of PLR routing table updates will be

PLR

⎥

⎦

⎥

⎢

⎣

⎢

IPT

⎥

⎦

⎥

⎢

⎣

⎢

PIP TT

, while the number of inter-PLR routing table updates will be

IPLR

= n -

⎥

⎦

⎥

⎢

⎣

⎢

IPT

. The expected cost for the PLR roaming algorithm will be:

{}

∑

∞

⋅+⋅+⋅=

)(

rPLRPLRIPLRIPLRHLRHLR

ROAM

npCNCNCNC

(1)

147

where C

HLR

is the cost of an HLR+ update, C

IPLR

the cost for inserting/updating an

inter-PLR link, C

PLR

is the cost for updating a routing entry, and p

(n) is the

probability that n different PLR servers are crossed within two calls. After the HLR+

is updated, the length of the path to the terminal consists of

PLR

⎥

⎦

⎥

⎢

⎣

⎢

⎥

⎦

⎥

⎢

⎣

⎢

−

PIP

(2)

PLR links (entries at the PLR routing tables), and

IPLR

= n -

⎥

⎦

⎥

⎢

⎣

⎢

PIP TT

– N

IPLR

(3)

Inter-PLR links. If C

is the cost for a direct terminal paging, O

PLR

is the overhead to

follow an entry in the PLR routing table and O

IPLR

the relevant overhead for the inter-

PLR roaming, the overall cost for the PLR paging algorithm will be

{}

rIPLRIPLRPLRPLRPAGE

CnpOLOLC ++⋅=

∑

∞

⋅

)(

(4)

In order to evaluate the p

(n), we assume that the mean rate λ of the call arrivals is

a Poison distribution and the interval between two PLR roaming instances is a random

variable, which for simplicity has a general density function described by a Gamma

distribution with mean 1/µ. The Laplace transform of the Gamma distribution is

rr CC

=⎯⎯→⎯

⎟

⎠

⎞

⎜

⎝

⎛

)(

γµλ

γµ

where p=λ/µ. For simplicity we have assumed an exponential distribution, thus γ=1. It

can be shown that (1) and (4) are equal to

1)1(1)1( −+

−

−+

−

PIPIP TT

IPLRr

IPLRPLRIPLR

ROAM

(5)

−+

−+=

1)1(

PIP

PIPIPLRIPLR

pPAGE

TTO

]1)1][(1)1[(

]1)1()1)[((

−+−+

−++−+−

IPPIP

PIP

PLRIPPLR

pOTO

(6)

Without the PLR algorithm the overall cost for maintaining the location information

and page the terminal is:

C +=

(7)

While the overall cost for the PLR architecture is

148

PAGEROAMPLR CCC +=

(8)

The roam (G

ROAM

), page (G

PAGE

) and overall (G

Total

) gains are

ROAM

G = ,

PAGE

G =

PLR

TOTAL

(9)

From (1)-(9), the

TOTALG

can be evaluated. If we assume that C

IPLR

= 2O

IPLR

and

PLR

= 2O

PLR,

from (8)-(9), we can depict the PLR roam, page and total gains as a

function of terminal Call-Mobility Ratio (p).

=2, T

=4, T

ROAM

PAGE

TOTAL

PLR

TOTAL

=2, T

=4, T

=2, T

=4, T

ROAM

PAGE

TOTAL

PLR

TOTAL

Fig. 4. PLR Algorithms Gain (C

PLR

= 0.45, C

IPLR

= 0.3)

ROAM

PAGE

PLR

TOTAL

PLR

TOTAL

PLR

TOTAL

PLR

TOTAL

=2, T

=4, T

=2, T

=4, T

Fig. 5. PLR Algorithms Gain (C

PLR

= 0.9, C

IPLR

= 0.6)

149

As shown in Fig. 4, the gain G

ROAM

of the PLR scheme can be up to 70%, while

the G

PAGE

leads to higher paging time. However, the overall gain G

Total

can be up to

60%. It should be underlined however that in this evaluation we do not measure the

actual G

PAGE

in case the system had to locate a terminal in heterogeneous adjacent

network location management systems. The graphs also show that as the terminal

Call-Mobility Ratio (p) increases, the G

ROAM

and the G

PAGE

gain degrease. When the p

is small, the user roams more often. This leads to more frequent updates and larger

paging paths, so smaller G

PAGE

The G

Total

increases as more updates are local, and the HLR+ is not informed so

often. If we increase the C

IPLR

and C

PLR

values, the gain of the overall PLR algorithm

degrades faster with large T

value, compared with small T

value (Fig. 5).

This is due to the fact that larger thresholds T

, T

lead to longer paths towards the

terminals, thus the system is more sensitive to the costs of inserting/updating a routing

entry in a PLR server.

5. Conclusions

Since a variety of mature wireless technologies are already available, a phased

approach may be deployed as evolving steps towards 4G. Future mobile terminals

will require to uninterruptedly roam from different in-building wireless networks, into

heterogeneous public picocellular/microcellular or even wide area macrocellular or

satellite networks.

Commercial public wireless LAN solutions however offer limited location

management capabilities compared to the traditional cellular networks. In order to

overcome these limitations, we introduced a Path Location Register (PLR) scheme for

Mobile Terminals Location Management. As has been shown in the performance

evaluation section, the proposed scheme reduces significantly the cost of mobile

terminal location update and paging, without dramatically increasing the system

complexity.

References

1. Th. Zahariadis, K. Vaxevanakis, Ch. Tsantilas, N. Nikolaou, N. Zervos, “Global Roaming in

Next Generation Networks,” IEEE Commun. Mag., Vol. 2, pp. 145-151, Feb.2002

2.] B.-N. Amotz, I. Kessler, M. Sidi, “Mobile users: To update or not to update?,” in Proc.

IEEE INFOCOM, vol. 2, June 1994, pp. 570–576.

3.] S. Tabbane, “Location management methods for third-generation mobile systems,” IEEE

Commun. Mag., vol. 35, pp. 72–84, Aug. 1997.

4. C. Perkins, “IP Mobility Support,” RFC 2002, Oct. 1996

5. A. Valko, “Cellular IP - A New Approach to Internet Host Mobility,” ACM Computer

Communication Review, January 1999

6. I.Akyldiz, W.Wang, “A Dynamic Location Management Scheme for Next-Generation

Multitier PCS Systems,” IEEE Trans. in Wireless Comm., Vol. 1, No. 1, pp.178-189, Jan.

2002.

7. A.Festag, H.Karl, G. Schaefer, “Current development and trends on handover design for All

IP wireless networks,” Technical University of Berlin, TKN-00-007, ver. 1.3, Aug. 2000

150

8. R. Ramjee, T.La Porta, L.Salgarelli, S. Thuel “IP-Based Access Network Infrastructure for

Next-Generation Wireless Data Networks,” IEEE Pers. Commun.,vol. 7, no. 4, Aug. 2000,

pp.34-41.

9. A. Campbell, J. Gomez, S. Kim, A. Valkó, C. Wan, Z. Turányi, “Design, implementation,

and evaluation of cellular IP”, IEEE Pers. Comm., vol. 7, no. 4, August 2000 pp. 42-49

10. Th. Zahariadis, N. Nikolaou, “Unified Wireless Access in Hot-Spot Environment”, IEEE

Communications Letters, Vol. 6, No. 6, pp. 259-261, June 2002

11. Y.Fang, I.Chalamtac, Y.Lin, “Portable Movement Modeling for PCS networks,” IEEE

Trans. Veh. Technolog., vol.87, no.8, pp.1347-1384, August 1999

151