10.1.1.59.2500

8/7/2019 10.1.1.59.2500

http://slidepdf.com/reader/full/1011592500 1/10IEEE Wireless Communications • August 200326 1536-1284/03/$17.00 © 2003 IEEE

P

WLAN

Intracellad hoc net

Advances in the

network architecture,

enhancements in the

signaling protocols,

provisioning of end-

to-end QoS, world-

wide seamlessmobility, and flexible

service provision are

among the major

research challenges

towards next-

generation wireless

networks.

(R)EVOL UT I ON TOWA RD

4G MOBI L E COMMUN I CA T I ON SY ST EMS

INTRODUCTION

With third-generation (3G) mobile networksalready launched in Japan and initial deploy-ment just beginning in Europe, it is an appropri-ate time to look beyond 3G and ask what thenext major development in mobile will be. It ispossible [1] to take the view that the progres-sion through the “generations” is simply aboutnew air interfaces and higher bandwidths. Thereis support in both Japan and the InternationalTelecommunication Union —Radiocommunica-tion Standardization Sector (ITU-R) for this asa definition of fourth-generation (4G) mobileas, potentially, a 100 Mb/s air interface. Analternative view [2] is that there will be an IPcore network, through which all traffic will flow,and a large number of access technologies, bothfixed and wireless, will provide connectivity.Concrete examples of these include digital sub-scriber line (DSL) and wireless LAN technolo-gies. Taking the example of the UnitedKingdom, it is estimated that there will be 5million DSL connections and 4000 WLAN

hotspots by 2005.

IP has fundamentally changed the nature of communications. It has become the de factostandard for data transmission and, with devel-opments in voice over IP (VoIP) such as Uni-versal Mobile Telecommunications System(UMTS) Release 5, is rapidly being used for allapplications. In part the impact of IP has beeneconomic, breaking apart the traditionally inte-grated fixed/mobile operator’s value chain andintroducing actors like Internet service pro-viders (ISPs) and content providers. One of the

reasons for this is that IP networks only trans-mit packets — services and intelligence arelocated at the network edge — and the primarydifferentiator in today’s IP networks is band-width. IP service creation is not integrated withnetwork service. The economic issues raised bythis are illustrated by the Japanese Imode sys-tem: users can either visit content within awalled garden of approved sites, or simply con-nect to the wider network.

The development of terminals, driven by aMoore’s law timescale, means that users willsoon have much more choice in how they con-nect to networks and obtain services (in thewidest sense). Smaller laptops, PDAs, and tabletPCs are converging with ever smarter mobilephones. The low cost of Bluetooth and 802.11a/bWLANs means that these will soon be incorpo-rated in a majority of these devices. Many com-mentators believe that ad hoc and personal areanetworks (PANs) will become commonplace asusers create networks within their homes, shareconnections, and form closed user groups. Inaddition, universities, enterprises, and hotspotowners are now able to deploy public accessWLANs across most of Europe; this activity willgreatly increase when 802.11h equipmentaccepted for use in the 5 GHz bands in Europeis available.

All these developments pose a fundamental

DAVE WISELY, BTEXACT

HAMID AGHVAMI, KINGS COLLEGE LONDON

SAMUEL LOUIS GWYN, LUCENT TECHNOLOGIES

THEODORE ZAHARIADIS, ELLEMEDIA TECHNOLOGIES

JUKKA MANNER, UNIVERSITY OF HELSINKI

VANGELIS GAZIS, NIKOS HOUSSOS, AND NANCY ALONISTIOTI, UNIVERSITY OF ATHENS

ABSTRACT

Advances in network architecture, enhance-ments in signaling protocols, provisioning of end-to-end QoS, worldwide seamless mobility,and flexible service provision are among themajor research challenges toward next-genera-tion wireless networks. The integration andinteroperability of all these technologies, alongwith new truly broadband wireless innovationsand intelligent user-oriented services will lead

toward the so-called 4G wireless networks. Inthis article we identify the key issues of an inno-vative transparent IP radio access system thattargets 4G networks.

TRANSPARENT IP RADIO ACCESS FOR

NEXT-GENERATION MOBILE NETWORKS

8/7/2019 10.1.1.59.2500

http://slidepdf.com/reader/full/1011592500 2/10IEEE Wireless Communications • August 2003 27

challenge to existing fixed and mobile telecomoperators. Users, faced with a range of accesstechnologies, an overcapacity of fiber backboneconnectivity, and services available from multipleproviders on the Internet, might well opt for thecheapest “bit carrier.” Optimal “anywhere” ser-vice provision, however, cannot be accomplishedin the absence of a number of compelling func-tions that can, when integrated into a package,

retain much more of the value chain for the net-work operator [3]:• Flexible and dynamic service creation and dis-

covery• Network reconfigurability management• Personalization and context awareness• Adaptation of services to access technologies

and terminals• Seamless access• End-to end quality of service (QoS)• Support for new network topologies such as

PANs and ad hoc networks• Billing and accounting services

In addition to these enhanced network func-tions, it is important for network operators tolower costs of IP packet delivery, especially forreal-time services such as voice, since this is theultimate purpose of the network. TransmittingIP packets over wireless links presents manyproblems, and current solutions are far fromoptimized. As an example, the common 802.11bWLAN technology is wholly unsuited to provid-ing a cellular voice service; the capacity forvoice-only services drops to less than 10 percentof that for data, there is no coordination withlayer 3 QoS signaling, encryption is inadequate,and a cellular WLAN network suffers high levelsof interference [4].

We believe the integration and interoper-

ability of all these diverse technologies, along

with new truly broadband wireless innovationsand intelligent user-oriented services will leadtoward a new generation of heterogeneouswireless networks, which has already beencalled 4G. In this article we propose an innova-tive transparent IP radio access system that tar-gets 4G networks. The article is structured asfollows. We describe a heterogeneous open net-work architecture that is able to fulfill potential

future service requirements in a cost-effectiveway. We identify the trends and research areasin the layers below the IP layer. We addresstwo critical issues in future mobile networks:user mobility and end-to-end QoS. The first isthe ability of the user to seamlessly roam in thenetwork, while keeping communication sessionsuninterrupted, while the second is the networkability to guarantee the agreed end-to-end qual-ity of the provided services. Users, however, donot care about the technological limitations, therestrictions of the wireless medium, and theheterogeneity of the network. Therefore, anintelligent middleware is described that aims tohide the network and technology details fromthe end user and provide truly transparent IPradio access.

OPEN ALL-IP NETWORK ARCHITECTURE

By definition it is difficult to make predictionsand foresee which network systems and serviceswill succeed in the forthcoming 10 years. Howev-er, one thing that can be expected is that sincenetwork operators have already made hugeinvestments in network infrastructure, 4G net-works are expected to integrate all heteroge-neous wired and wireless networks, and provideseamless worldwide mobility [5].

An abstract view of the envisaged network

I Figure 1. Open all-IP network architecture.

Satellitecoverage

Node B

RNC

2G+/3Gaccess network

Ad hoc/PANmoving network

Police

BTS

Accesspoint

Accesspoint

Accessrouter

Access routerconcentrator

Wireless LAN

BSC

SGSN

GGSN

PSTN/ISDN/fixed wirelessnetwork

SIPHLRAAA

GSM/GPRSaccess network

Media gatewayaccess router

IP core network/ Internet

Node B

8/7/2019 10.1.1.59.2500

http://slidepdf.com/reader/full/1011592500 3/10IEEE Wireless Communications • August 200328

architecture is shown in Fig. 1. The 4G core IPnetwork will be dominated by connectionless IPpacket switching technologies and interface viaaccess routers with multiple wireless access tech-nologies like indoor (e.g., IEEE 802.11a/b,HiperLAN/2, Bluetooth) and cellular 3G/4Gnetworks. Gateway access routers will interfacethe network with a multitude of wired a nd airinterfaces, inherited from earlier-generationcommunication systems like the public switchedtelephone network/integrated services digital

network (PSTN/ISDN) and 2G/2G+ technolo-gies: Global System for Mobile Communications

(GSM), Digital Cellular Service (DCS), GeneralPacket Radio Service (GPRS), and code-divisionmultiple access (CDMA). Unified mobility man-agement will enable common handing of cus-tomer profiles and Internet addresses, userlocation, and authorization, authentication, andaccounting (AAA) functions, facilitating seam-less handover between multiple operators’ wire-less networks.

Distributed ad hoc, spontaneous, and in manycases moving networks, deployed in the unli-

censed bands, will instantaneously provide wire-less personal area networks (PANs). A PANnetwork may include any collection of devicesthat belong to or are carried by a ne tworkeduser (e.g., cell phone, laptop, earphones, GPSnavigator, Palm Pilot, beeper) and form his/herpersonal “PAN bubble.” The bubble can expandor shrink dynamically depending on a user’senvironment and needs. Such access is importantwhen the mobile user enters a new location andaims to quickly sense and control the environ-ment (e.g., gain access/connectivity, control thetemperature, adjust the lighting) or be recog-nized by the environment (e.g., welcome mes-

sage, uninterrupted communication, automaticselection of background music).Apart from nodes, mobility is also foreseen

for entire subnetworks, which change their pointof attachment to the global Internet. An exam-ple of a moving network is the PAN of a userwho is inside a moving car, or is a passenger ona train or ship that connects his/her PAN to theInternet via the company’s private network.Moreover, mobile proxy/ad hoc access points(e.g., the car in Fig. 1) may provide bridgingoperations to devices that belong to the PAN,but may not have the means to directly commu-nicate with the rest of the IP network. Combin-ing the notion of mobile or moving networks

with the security and privacy provided by virtualprivate networks (VPNs), a new type of servicecan be introduced, called Mobile VPNs.

In order to support this heterogeneous multi-functional network architecture, the IP-basedprotocol stack of Fig. 2 is proposed. At the phys-ical layer, an innovative multitechnology, airinterface is assumed, able to provide connectivityto existing cellular (2G/3G), WLAN (802.11a/b/h, HiperLAN/2), wireless PAN (Bluetooth)and future (4G) wireless networks. Apart fromadaptability in the physical layer, the proposedwireless adaptive data link control (DLC) willprovide mechanisms to interface to the networktechnology chosen and to provide information tothe upper layers. Over the DLC, various IP pro-tocols will provide the means of network com-munications including transportation, signaling,messaging, QoS, mobility, security, control, andmanagement. Finally, a transparent middlewarelayer will provide for an abstract v iew of theunderlying mobile network infrastructure andfacilitate focused application development.

SUB-IP INNOVATIONS

The aim of future radio access is to ensure theavailability of a transparent wireless transportand delivery mechanism that permits the distri-

bution and creation of services in an IP-like

I Figure 2. The proposed IP-based protocol stack.

Applicationlayer

IP networklayer

Sub-IPlayer

Applications

Application middleware

IP control andmanagement:COPS, SNMP,

RSVP,ICMP

IP networking:IPv4, IPv6, DNS, DHCP, zConf, multicast, multihoming

Wireless adaptive DLC

WPAN WLAN 2G/3G 4G

IP middleware:SIP, SLP, IMPP, ...

IP messaging:SMPT, HTTP, BEEP, CPIM, ...

IP transport:TCP, UDP, DCCP, SCTP, RTP

IP mechanisms:QoS, mobility, security

I Figure 3. Macro service delivery via direct and indirect wireless access.

PAN

PAN

PAN

4GWLAN

Core IP(future Internet)

UTRAN

Servicedelivery

Servicedelivery

Intracell PANad hoc network

8/7/2019 10.1.1.59.2500


manner, with reliability matching that of wirelinenetworks. Moreover, its architecture must ensurethat the cost of wireless data delivery is effectiveregardless of the access network used. The abso-lute need for cost effectiveness will ensure thatthe environment for attractive well priced dataservices is at least in place and that competitiveadvantages to an operator can be maintainedwhile allowing the end user a rich set of compet-itive data services.

In order to achieve cost effectiveness, future

wireless infrastructures have to integrate trans-parently multiple heterogeneous wired andwireless radio access networks that have alreadybeen deployed, supported by a full IP-basedcore network. As shown in Fig. 2, 2G/3G+ cel-lular, WLAN, PAN, and ad hoc PAN accessnetworks will comprise this heterogeneous envi-ronment.

In the envisaged scenario, the traditional (adhoc) PAN network concept is extended from acomposition of devices that form ad hoc connec-tions between themselves to a composition of devices that have cellular and wireless accesscapabilities. In this respect, the PAN takes on

the additional property of virtual terminal equip-ment(VTE), and the user may be connected,simultaneously or in a sequential manner, to thecore network via a variety of wireless accessmechanisms, while maintaining intracell connec-tivity with other users as a participant node of anad hoc network. This connectivity richness of theextended VTE concept may further enhance theservices exploitation potential. For example (Fig.3), because of the ad hoc capability of the PAN,an individual user may be receiving a servicedirectly via the individual’s PAN VTE connec-tion to a wireless access medium, but also via anad hoc connection through a number of adjacentPAN VTEs over which the user has temporary

access rights. In this sense the PAN VTE is alsoacting as a proxy and ad hoc access point (PA-AP) to the wireless access network. In effect thePA-AP acts as a proxy gateway for the wirelessaccess network so that the service layer is able toshare, receive, and create mobile services. Thecharacteristics of these networks are given inTable 1.

In the wireless service delivery example of Fig. 4, the wireless access network treats theVTE PAN as a s ingle terminal entity. In thisenvironment, the research challenge is toensure the maintenance of sufficient band-width, with low latency, to the PAN VTE via avariety of access mechanisms, while the PANVTE acts as PA-AP for macro service deliverygiving the impression of wire-like quality. Thechallenge from the mobile PAN environment isto provide enough interference mitigationstrategies, so that the internal PAN VTE andPA-AN environments can be maintained flexi-bly. An example of this challenge is to havemacro service delivery achieved by a variety of enhanced cellular systems either simultaneouslyor sequentially.

A key feature to achieve this wireless servicedelivery environment lies in the flexibility it pro-vides to the higher layers so that transparencycan be achieved. This flexibility is attained via

the application of polymorphic concepts to the

radio access layers. In this regard, the aim is toreduce the complexity of the system to the high-er layers using techniques that apply the invoca-tion of multiple methods via a single interface;these methods are bound to a given air interfaceat runtime. It is this single interface, multiplemethods, and late binding techniques that giveintelligence to the radio access layers and theirpolymorphism.

Apart from adaptability in the physical layer,it is proposed that wireless DLC will have mech-

anisms to interface to the network technology

I Table 1. Definitions and characteristics.

Name Definition and characteristics

Ad hoc PAN No infrastructure (private network)• Dynamic topologies• Inter-PAN connection (device to device of different PANs

using the same interface)• Inter-PAN gateways (different standard relay)

Cellular Infrastructure-based• Operator-owned (public network)• Large range

• Low/medium data ratesPA-AP Proxy and ad hoc access point

• This is a VTE element that can form ad hoc connectionsbetween VTEs

PAN No infrastructure (private network)• Virtual terminal consisting of various devices, all possibly

with at least one access interface• Low range• High data rates• Dynamic topologies• Any number of gateways to infrastructure-based networks

VTE Virtual terminal equipment• From a cellular perspective

• Forms a single manageable entity• Composed of PAN wireless access elements

WLAN Infrastructure-based• Not necessarily operator-owned (public/private network)• Low range• High data rates

I Figure 4. A wireless service delivery environment.

PAN

PAN

PAN

PAN ad hocnetwork

PAN track

accrosscellularsystem

PAN interferencemitigation zone

PAN

PAN

PAN

PAN

Global cellularsystem

8/7/2019 10.1.1.59.2500


chosen and provide information to the upperlayers. Moreover, in cooperation with the mid-dleware layer, it will contain procedures formapping several control issues of the network toDLC control information. These procedures mayrefer to special messages such as connectionsetup or connection release at the network level,which have to be translated to the DLC level.The DLC layer will support QoS over the wire-less link for each connection. Thus, the parame-ters of the network connection have to betransferred to the DLC level to serve the DLCconnection accordingly.

MOBILITY MANAGEMENT

Next-generation mobile networks should providemechanisms that enable seamless roamingbetween multiple heterogeneous subnetworks[6]. In more detail, 4G wireless networks shouldprovide:

Personal mobility. This is the ability of theuser to access his/her personalized network ser-vices while away from his/her home network.The proposed framework aims to provide usermobility based on an intelligent end-to-end loca-tion-aware middleware layer.

Terminal mobility. This is the ability of thenetwork to maintain continuous services evenwhen the user’s terminal changes locations. Ter-minal mobility becomes more challenging forstreaming applications with strict QoS require-ments and devices with multiple heterogeneousradio access technologies. Handing over multi-homed terminal mobility will require intertech-nology handovers, and resource managementinitiated by layer 2 triggers and supported byenhanced signaling protocols.

Network mobility is the ability of the networkto support roaming of an entire subnetwork,structured or ad hoc. In this case the core andaccess networks should be able to recognize themoving network as a special roaming node.

Mobile IP, which initially addressed the

mobility management problem, introduces signif-icant network overhead in terms of:• Increased delay due to triangular routing• Packet loss due to registration delays• Increased signaling due to frequent terminal

registrationRoute optimization can improve performance,but cannot provide acceptable service quality,especially for streaming applications like VoIP.The network overhead is drastically increased if mobile hosts frequently change their point of attachment. Based on the assumption that themajority of frequent handovers take placebetween nodes of the same regional network,

many IP micromobility protocols have been pro-posed aimed to provide fast seamless localmobility. IP micromobility protocols reduce theround-trip delay and eliminate the end-to-endregistration signaling overhead by handling localmovements of the mobile node transparentfrom the global Mobile IP network. This isachieved by maintaining regional locationdatabases that map mobile node identifiers tolocation information, which is used for localdelivery of the packets. In order to furtherreduce registration signaling, IP paging tech-niques are introduced. Just like cellular net-works, micromobility protocols take into accountthe operational mode of the mobile node andmaintain approximate location information forthe idle users. In this way, intermediate signal-ing messages generated by frequent local move-ments of an idle node are suppressed, with thecost of increased delay when a new call is initi-ated. In general, micromobility managementprotocols [7] may be categorized into two groups(Fig. 5).

The hierarchical tunneling approach: Theprotocols that fall in this category incorporate anumber of foreign agents (FAs) in a tree-likestructure. Each FA maintains part of a locationdatabase. The entries of the database containthe address of the next FA in the path toward

the mobile node. Data packets that address a

I Figure 5. a) Hierarchical tunneling and b) explicit terminal routing approaches.

(a) (b)

Routing information Foreign agent (FA) Routing node

TunnelEdgenode

FA

Routing pathRootFA

FAFA

FA

Coremobile IPnetwork

Coremobile IPnetwork

IP Micromobility

protocols reduce the

round-trip delay and

eliminate the

end-to-end

registration signalingoverhead, by

handling local

movements of the

mobile node

transparent from the

global Mobile IP

network.

8/7/2019 10.1.1.59.2500


specific terminal are delivered encapsulated tothe root FA of the visiting tree network. EachFA in the lower layers of the tree structure thatis the endpoint of a tunnel has to decapsulatethe packets, search its location database for acorresponding entry, and then re-encapsulateand forward them through another tunnel downto the terminal’s point of attachment. When theterminal moves, it propagates location updatemessages up to a tunnel endpoint, which islocated in both the old and new paths. Exam-

ples of micromobility protocols that use thehierarchical tunneling approach are Hierarchi-cal Mobile IP, Hierarchical Mobile IPv6, andthe Brain Candidate Mobility Protocol(BCMP). Hierarchical tunneling micromobilityprotocols may have to introduce newcontrol/signaling messages or require a hierar-chical tree structure.

The explicit terminal routing approach:The protocols that fall in this category avoidthe overhead introduced by encapsulation andtunneling, with the cost of introducing explicitrouting entries for each terminal. Instead of location database, each node keeps a (kind of)

routing table that contains an index to thenext hop node in the path towards the termi-nal’s point of attachment. When a terminalmoves, location update information is propa-gated explicitly (via signaling messages) orimplicitly (via data message snooping) up toan anchor or crossover node, which is locatedin both the old and new paths or up to theroot FA. Examples of micromobility protocolsthat use the explicit terminal routing approachare Cel lular IP, Handoff Aware WirelessAccess Internet Infrastructure (HAWAII), andUnified Wireless Access (UniWA). Also inthis category fall protocols like “iwander,”which are based on Ethernet switching and

layer 2 MAC to port number binding. Theseprotocols do not require a hierarchical net-work structure, but may face scaling limita-tions due to the increased amount of routinginformation especially at nodes closer to theedge node.

In the last few years, there has been consider-able debate in IETF on suitable fast and seam-less handoff extensions. However, the 4Gnetwork architecture proposes enhancements inthe existing signaling and transport protocols. Inthe forthcoming years, special focus is expectedon many research areas, including:• Scaling over large networks (e.g., large number

of terminals, large number of cells per accessrouter)

• Signaling overhead vs. response delay time(layer 2 vs. layer 3 signaling vs. control proto-cols)

• Location update vs. IP paging• Roaming information (terminal velocity/direc-

tion/moving model/profile vs. open sessionrequirements/contract vs. network resources/congestion)

END-TO-END QOS

Another key differentiation of next-generationmobile networks will be the provisioning of

end-to-end QoS for interactive stream-based

multimedia services (e.g., video telephony,videoconference). Currently two ways to pro-vide differentiated treatment of flows may beidentified: via differentiated services codepoints (DSCPs) or application signaling. In theDiffServ-based solution an end host can markthe transmitted packets with proper DSCP val-ues to trigger certain service responses. Howev-er, as the DSCP values define only relativepriority and their value may not persist end toend, various problems may arise in their inter-

pretation in intermediate nodes. Thus, thismethod cannot guarantee end-to-end QoS pro-visioning. On the other hand, various QoS sig-naling protocols have been proposed. TheResource Reservation Protocol (RSVP) andIntegrated Services specifications are amongthe most widely accepted. However, none of these protocols are foreseen to provide end-to-end QoS in all situations in tomorrow’s hybridnetworks [8]. The IETF Next Steps in SignalingWorking Group is currently defining require-ments and architectures for IP signaling, evalu-ating existing signaling protocols, and lookingat new developments. One application of a new

signaling protocol is QoS.The major problem in providing QoS in amobile environment is the coupling of QoS andmobility management; that is, how to keep pro-viding the same level of quality to the packetflow during and after a handover. Mobile IP-based networks require the same level of mobility support as GSM networks, which offervoice and data services, but an ongoing call isvery rarely dropped due to movement. Howev-er, no technology currently exists that wouldallow fully seamless IP mobility and still becommercially feasible. The question is alwaysabout a balance between resource usage andlevel of service.

There have been several designs for exten-sions to RSVP to allow more seamless mobility.One solution couples RSVP and mobility man-agement mechanisms, and proposes small exten-sions to localize the RSVP signaling [9]. Theextension allows the mobile host to be in chargeof both upstream and downstream reservationsand re-set up reservations on both directionsimmediately following a handover. Also, contexttransfers studied in the IETF Seamoby WorkingGroup aim at enhancing and minimizing signal-ing needed during handoffs. Another group of examples are protocols based on advance reser-vations, where neighboring access points keepresources reserved for mobile nodes moving totheir coverage area. When a mobile noderequests resources, the neighboring accesspoints are checked too, and a passive reserva-tion is done around the mobile node’s currentlocation. The problems with these schemes arethat they require topological information of theaccess network and usually advance knowledgeof the handover event. Furthermore, the waythe resources reserved in advance are used inthe neighboring service areas is still an openissue [10].

What is missing is a common QoS signalingframework, one that application designers coulduse to make their applications QoS-aware. Cur-

rently, there exist too broad a range of QoS

Another

key differentiation

of next-

generation mobile

networks will be the

provisioning ofend-to-end QoS for

interactive stream-

based multimedia

services (e.g.,

video telephony,

videoconference).

8/7/2019 10.1.1.59.2500


mechanisms, for both signaling and provisioning,which makes it impossible for an applicationdesigner to decide on a technology; moreover,the very same signaling protocol must be sup-ported by the network too. A flexible way to

make applications QoS-aware is to use middle-ware that can map the applications’ QoSrequests to QoS technology used at the IP layer.This is discussed in more detail in the next sec-tion.

Still, at the end of the day, the deployment of various QoS mechanisms is a classic “chickenand egg” problem. We need new business mod-els before any operator is willing to invest inexpensive new QoS mechanisms. Take, for exam-ple, GPRS: the specifications included a numberof different QoS parameters (e.g., bandwidthsand priorities), but those are not currently usedanywhere.

On the other hand, we also need a clearpath in deploying QoS in IP networks incre-mentally, network by network, application byapplication. Currently, most parts of the tech-nologies needed for providing distinguished ser-vices in a best-effort IP network are alreadyavailable. An architecture that combines per-application signaling and traffic classes withinthe backbones would be a very interesting solu-tion [11]. There are still some open issues, butthe enabling technologies for QoS-enabled ser-vices are already here. The question is, whichoperators make the first move and decide on amigration path from best effort to the integrat-ed services network foreseen by the IETF in

the mid 1990s?

MIDDLEWARE AND SERVICES

Typical application development practices in theproposed environment become very complicatedby the requirement to anticipate all possible net-

working contexts during application design andto interpolate network-dependent code withinthe application’s core logic. It is therefore neces-sary to introduce a middleware architecture thatabstracts the technological details of the mobilenetwork infrastructure and facilitates applicationdevelopment, deployment, and management. Ingeneral, the term middleware denotes a softwarelayer that resides between the different parts of a distributed application in order to shield themfrom the heterogeneity of the underlying plat-forms. In the case of next-generation mobile ser-vice provision, the role of middleware is notrestricted to handling fundamental (albeit impor-tant) tasks like seamless “binding” and commu-nication between distributed applicationcomponents; it should also incorporate higher-level functionality to facilitate adaptable serviceprovision. Furthermore, deployment and use of diverse distributed applications may require pro-visioning — or even reconfiguration — of net-work equipment in multiple administrativedomains, an issue best tackled by service provi-sion platforms that can act as a control point forservice provision flexibility and adaptability.

In a heterogeneous infrastructure applica-tions face an enormous range of operationalparameters, thereby making adaptation a criticalissue. Adaptability requires constant awareness

of the evolving service provision context (e.g.,

I Figure 6. The proposed architecture.

Open interface Message queue Functional feature Protocol adapter Reused technology

Discovery

Framework

Naming

Notification

Session QoS Security

Rad

ius

Diam

eter

MAP

BCMP

Mobile IP

SNMP

CO

PS

MP

LS

RSVP

DiffServ

CAM

EL

HT

TP

SIP

Events Provisioning

Deployment

B

illing

Accoutning

Service

provision

platform

Sec

urity

Independentapplicationdeveloper

Third-partyapplicationprovider

Value-addedservice

provider

Profiling Context

Service support layer

Messaging transport

Network abstraction layer

Persistence PolicyAccount

management

Registrationmanagement

Policymanagement

Capabilitiesmanagement

Configurationmanagement

Application

Application

Reused focus

8/7/2019 10.1.1.59.2500


user preferences, user location and status, mobileterminal capabilities). A system is context-awareif it explores context to provide relevant infor-mation and/or services to the user, where rele-vance depends on the user’s current tasks andpreferences. Thus, context awareness is regardedas the ability to autonomously adapt, eitherproactively or reactively, to observed orannounced changes in the environment (e.g.,changes in the user’s location or the mobile ter-minal’s power drain rate). Consequently, middle-

ware must support a flexible context definitionscheme and also provide mechanisms for contextcapture, collection, management, and distribu-tion to concerned applications using open inter-faces that impose no restrictions on applicationarchitecture or implementation. Middleware ser-vices must hide the implementation details of underlying networks while exposing particularcontextual aspects (e.g., available QoS classes,resource status) important for applications tobehave efficiently and adapt when necessary.

Adaptation strategies may also be applied tofunctional network elements in order to dynami-cally adjust their grade of service according to

specific application criteria. For instance, in thecase of a multimedia streaming applicationwhose bandwidth requirements cannot be met bythe current wireless bearer service, the possibleadaptation strategies are:• To lower the video and audio codec resolu-

tions• To negotiate higher bandwidth for the bearer

service• A combination of bothThe application of such two-way adaptationstrategies in heterogeneous wireless environ-ments requires a network abstraction layer thatexposes the common functional features of underlying mobile networks to the middleware

intelligence in a technologically agnostic way. Byusing open interfaces, mobile networks of differ-ent technologies can be “plugged in” to the mid-dleware and make their signaling capabilitiesavailable to applications.

An important aspect of contextual informa-tion concerns user preferences. The middlewaremust enable the retrieval and handling of userpreferences in a generic way so that applicationsor even the middleware itself may align theiradaptation strategy to the optimal trade-off between network QoS, application presentationfidelity, communication security, and overall ser-vice cost [12]. Since users may formulate arbi-trarily complex preferences that must betranslated to reconfiguration actions on mobilenetwork elements without jeopardizing theirintegrity, flexible and generic configuration man-agement emerges as a paramount issue. Figure 6below illustrates the main layers and compo-nents of the proposed policy-enabled middle-ware architecture, highlighting the majorresearch issues.

The framework enforces security, integrity,and access policies by acting as an initial refer-ence point for all access to and subsequent useof middleware services through its discoveryinterface. Parlay and OSA also adopt a frame-work-based architecture. Furthermore, the

framework coordinates complex procedures

(e.g., policy and configuration management) thatrequire interactions across different layers of themiddleware architecture.

The service support layer provides a set of open interfaces categorized according to majorapplication concerns, such as contextual aware-ness, notification requirements, as well as policy,profiling, and persistence services. Applicationcomponents may discover and use these inter-faces in a technologically neutral way via theframework.

The network abstraction layer provides a setof generic reusable functional blocks that plug into the middleware architecture, providing an up-to-date view of the collective capabilities of theunderlying wireless networks, whether cellular,satellite, ad hoc, PAN, M-VPN, or any similarcombination. In addition, they export open inter-faces that allow higher layers to use selectedinfrastructure functionality, including retrieval of network state information and triggering of reconfiguration actions.

Messaging transport provides message-basedtransport services for all middleware compo-nents, thereby supporting asynchronous commu-

nications and facilitating loose coupling of interacting components. Ideally, the actual net-work protocols employed for information trans-fer would be dynamically selected according tocontextual information.

The service provision platform exploits mid-dleware capabilities through the framework tosupport secure and flexible service deploymentfor third-party-developed applications and value-added services. Furthermore, it engages duringservice discovery to customize the list of applica-ble services according to mobile terminal capa-bilities, wireless system properties (e.g., trafficload), and user privacy preferences.

In summary, the middleware architecture

defines a set of functional capabilities and openinterfaces that can be used as building blocks forinnovative application development in next-gen-eration mobile networks. By using the middle-ware API and interfaces, operator- andthird-party-developed applications will be able todiscover user preferences, be informed of specif-ic events in the network infrastructure (e.g.,handoff), and make informed adaptation deci-sions (e.g., content adaptation) when necessary.These middleware services are further exploitedand integrated into the service provision plat-form facilities for independent (third-party)application providers.

CONCLUSIONS

In the future mobile networks will display amuch richer topology than today’s 2G and 3Gintegrated systems: users will create personalarea networks, they may have multiple connec-tion options, and mobile VPNs will be dynami-cally modified and created. We believe that IPconnectivity will form the basis of the network-ing functionality of such networks, providingpacket delivery, QoS, and mobility support; andwe have developed an outline IP architecture forsuch a system. Given the open nature of IP, net-work operators, we believe, have two main goals:

reducing the costs of delivering IP packets to

The middleware

architecture defines a

set of functional

capabilities and open

interfaces that can

be used as buildingblocks for innovative

application

development in next-

generation mobile

networks.

8/7/2019 10.1.1.59.2500


mobile users and offering access to common ser-vices over any access technology through func-tions such as service discovery, payment, andadaptation.

In this article we present concepts, such as anadaptive data link control layer and reconfig-urable radio, that should enhance the efficiencyof IP multimedia service delivery over wirelesslinks and help support users connecting throughnew or extended network domains not con-trolled by their network provider.

Also presented are the trends in mobility andQoS frameworks for the mobile access network,highlighting the need for a common QoS signal-ing framework to support end-to-end QoS overthese heterogeneous networks and for furtherdevelopments of an intergraded, and hence effi-cient, QoS mobility solution. One of the majoroutstanding issues is that of securing QoS signal-ing, since QoS is a valuable network resource;this is particularly acute for handovers where ahigh level of signaling between mobile nodes andaccess routers is required. This issue is unre-solved and will potentially lead to considerabledelays in completing handover.

Current middleware technologies are notsuited to mobile environments; seamless accessrequires the adaptation of services as dictated byusers’ profiles. Multimedia service may berequired to be delivered over “complex” connec-tions, where, for example, voice and video travelover different access paths. It is the job of themiddleware to abstract the underlying networkservices and performance. We have developed amobile middleware architecture to support theconcept of seamless access, breaking the func-tionality down into service support, networkabstraction, and messaging layers.

ACKNOWLEDGMENT

This work has been performed in the frame-work of the TIPRA consortium. The authorswould like to acknowledge the contributions of their colleagues from BTexact, Telefonica ID,Portugal Telecom, TeliaSonera, NTT,TML(NEC), Fujitsu, Mitsubushi, Alcatel, Tele-tel, Qualcomm, Lucent Technologies, Thales,TTI, U4EA, Ellemedia, Alphamicro, ITRI,CEA-LETI, Agora, Kings College London,University of Surrey, University of Bristol,UPC, University of Murcia, Toshiba, Universityof Athens, UPM, University of Helsinki, Uni-versity of Singapore, and CEA.

REFERENCES[1] N. Nakajima and Y. Yamao, “Development of 4th Gen-eration Mobile Communication,” Wireless Commun.and Mobile Comp., vol. 1, no. 1, Jan. 2001.

[2] J. Pereira, “Fourth Generation: Now, It Is Personal,”Proc. 11th Int’l. Symp. Pers., Indoor and Mobile RadioCommun., 18–21 Sept. 2000, vol. 2, pp. 1009–16.

[3] N. Houssos et al., “Value Added Service Management in3G Networks,” NOMS ’02, Florence, Italy, Apr. 2002.

[4] L. Burness et al., “Wireless LANs — Present and Future,”to be published in BTTJ Special Issue on Mobile Sys-tems, Sept. 2003.

[5] T. Zahariadis, “Trends in the Path to 4G,” IEE Commun.Eng. Mag., Feb. 2003, pp. 12–15.

[6] T. Zahariadis et al., “Global Roaming in Next-GenerationNetworks,” IEEE Commun. Mag., vol. 40, no. 2, Feb.2002, pp. 145–51.

[7] A. Cambell et al., “Comparison of IP Micromobility Proto-

cols,” IEEE Wireless Commun., Feb. 2002, vol. 1, pp. 72–82.

[8] J. Manner and X. Fu, “Analysis of Existing QoS Sig-nalling Protocols,” IETF NSIS WG, Feb. 2003, draft-ietf-nsis-signalling-analysis-01.txt.

[9] J. Manner and K. Raatikainen, “Extended Quality-of-Ser-vice for Mobile Networks,” IEEE/IFIP IWQoS 200), Karl-sruhe, Germany, June 6–8, 2001; Springer LNCS Series,vol. 2092, pp. 275–80.

[10] B. Moon and H. Aghvami, “RSVP Extensions for Real-Time Services in Wireless Mobile Networks,” IEEE Com-mun. Mag., Dec. 2001, pp. 52–59.

[11] Y. Bernet et al., “A Framework for Integrated ServicesOperation over Diffserv Networks,” RFC 2998, Nov. 2000.

[12] N. Alonistioti, N. Houssos, and S. Panagiotakis, “AFramework for Reconfigurable Provisioning of Services

in Mobile Networks,” ISCTA 2001, Ambleside, U.K.,2001, pp. 21–26.

ADDITIONAL READING

[1] C. Perkins, “IP Mobility Support,” Internet RFC 2002,Oct. 1996.

[2] A. Festag, H. Karl, and G. Schafer, “Current Develop-ment and Trends on Handover Design for All IP Wire-less Networks,” Tech. Univ. of Berlin, TKN-00-007, v.1.3, Aug. 2000.

[3] R. Ramjee et al., “ IP-Based Access Network Infra-structure for Next-Generation Wireless Data Networks,”IEEE Pers. Commun., Aug. 2000, pp. 34–41.

[4] T. Zahariadis and N. Nikolaou, “Unified Wireless Accessin Hot-Spot Environment,” IEEE Commun. Lett., vol. 6,no. 6, June 2002, pp. 259–61.

[5] C. Keszei et al., “Evaluation of the BRAIN Candidate

Mobility Management Protocol (BCMP),” IST GlobalSummit 2001, Barcelona, Spain, 2001.[6] A. Campbell et al., “Design, Implementation, and Evalu-

ation of Cellular IP,” IEEE Pers. Commun., Aug. 2002,pp. 42–29.

[7] D. Wisely et al., IP for 3G, Wiley, 2002.[8] IST EU Project BRAIN del. 2.2, “BRAIN Architecture

Specifications and Models, BRAIN Functionality andProtocol Specification,” http://www.ist-brain.org

[9] IST EU Project MIND (Mobile IP Network Development)del. 1.2, “Top-level Architecture for Providing SeamlessQoS, Security, Accounting and Mobility to Applicationsand Services,” http://www.ist-mind.org

BIOGRAPHIES

DAVE WISELY ([email protected]) has worked for BT for15 years in the fields of networks and mobility research.He pioneered optical wireless free-space links in the early1990s, constructing a 4 km 1500 nm system across centralLondon using optical amplifiers. He has worked in the fieldof mobility for the past eight years, looking first at wirelessATM and HIPERLAN type 1 systems, and more recently intoWLAN and cellular mobile systems. He was one of the pio-neers of an all-IP solution for future developments of 3G.He currently heads up BT’s UMTS and 4G research unit andwas also technical manager for the influential EU ISTBRAIN/MIND project. He has just published a book entitledIP for 3G and is currently leading an initiative on 2GWLANs providing QoS and handover over campus-sizednetworks.

HAMID AGHVAMI [SM] obtained his M.Sc. (1978) and Ph.D.(1981) degrees from King’s College, University of London.He continued as a postdoctoral research associate workingon digital communications and microwave techniques pro-jects sponsored by EPSRC, joining the academic staff in

1984. He is presently director of the Center for Telecom-munications Research at King’s. He does consulting on dig-ital radio communications systems for British andinternational companies. He has published over 230 techni-cal papers and lectured worldwide. He was a ComSoc dis-tinguished lecturer, and a member, chair, and vice-chair oftechnical programs and organizing committees of manyinternational conferences. He chaired the CommunicationsChapter of the U.K. and Republic of Ireland (1990–1998).He is a Fellow of the IEE.

SAMUEL LOUIS GWYN served in the Royal Navy from 1981 to1991 as a nuclear reactor specialist. On leaving the RoyalNavy he went on to study at Queen Mary and WestfieldCollege, University of London, receiving an M.Eng. in com-munication engineering in 1995. Further studies at QueenMary and Westfield College led to a Ph.D. in the applica-tion of nonlinear fynamics to teletraffic modeling in 1999.

Since November 1998 he has been with the Global Wire-

In the future mobile

networks will display

a much richer

topology than

today’s 2G and 3G

integrated systems:users will create

personal area

networks, they

may have multiple

connection options,

and mobile VPNs will

be dynamically

modified and

created.

8/7/2019 10.1.1.59.2500

http://slidepdf.com/reader/full/1011592500 10/10IEEE Wi l ss Comm ic tio s • A st 2003 35

less Systems Research Department of Bell Laboratories,Lucent Technologies, United Kingdom, where he is involvedin the development of advanced protocols and networkarchitectures for wireless communications. His currentresearch interests include nonlinear dynamics, complexitytheory, agent-based systems, software architectures andinfrastructures, software protocols, 4G systems, mobility,and resource management.

THEODORE B. ZAHARIADIS ([email protected]) receivedhis Ph.D. degree in electrical and computer engineeringfrom National Technical University of Athens, Greece, andhis Dipl.-Ing. degree in computer engineering from theUniversity of Patras, Greece. Since 1997 he has been with

Lucent Technologies, first as technical consultant to ACT,Bell Laboratories, subsequently as technical manager ofEllemedia Technologies (affiliate of Bell Laboratories),Athens, Greece, and since November 2001 as a technicaldirector. At present, he is leading R&D of multimedia plat-forms and 3G core network services. He is responsible forthe development of QoS multimedia service platforms,home network systems, and wireless protocols implemen-tation. His research interests are in the fields of broadbandwireline/wireless/mobile communications, interactive servicedeployment over IP networks, embedded systems, and resi-dential gateways. He has authored more than 30 papers,and served as principal guest editor of many special issuesof magazines and journals.

JUKKA MANNER received his M.Sc. in computer science witha minor emphasis in industrial management from the Uni-versity of Helsinki in 1999 and a Ph.Lic. in 2002. He has

just completed his Ph.D. thesis on IP QoS in mobile envi-ronments for public defense. His research focuses onmobile and wireless communications, and QoS in future-generation mobile networks. He has participated in severalnational research projects (funded by the National Technol-ogy Agency of Finland and industry) and in EC projectsBRAIN (IST-1999-10050) and MIND (IST-2000-28584). Hehas actively participated in the IETF, for example, in work-ing groups Seamoby, NSIS, and TSVWG. He has publishedover 20 articles in conferences and journals.

VANGELIS GAZIS ([email protected]) received his B.Sc. andM.Sc. (communication networking) from the Departmentof Informatics at the University of Athens, Greece, in 1995and 1998, respectively, and his M.B.A. from the AthensUniversity of Economics and Business in 2001. From 1995until now, he has been with the research staff of the Com-munication Networks Laboratory (CNL) at the Department

of Informatics in the field of mobile ad hoc networks andcellular systems (MOBIVAS, ANWIRE). In parallel, he workedwith a number of established companies in the IT sector asconsultant. He is currently pursuing a Ph.D. in the Depart-ment of Informatics. His research interests include flexibleand adaptable service provision, management of reconfig-urable systems and services, quality of service, billing, andbusiness model issues in 3G/4G mobile networks.

N IKOS HOUSSOS ([email protected]) received a B.Sc.degree from the Department of Informatics at the Universi-ty of Athens in 1998 and an M.Sc. (with distinction) intelematics (communications and software) from theDepartment of Electronic and Electrical Engineering, Uni-

versity of Surrey, United Kingdom, in 1999. Since 1999 heis pursuing a Ph.D. at the Department of Informatics andTelecommunications, University of Athens. He is also a staffmember of the Communication Networks Laboratory ofthe University of Athens. He is involved in projects MOBI-VAS (including workpackage leadership), ANWIRE, andPOLOS of the European Union IST framework. His mainresearch interests include middleware platforms and busi-ness models for beyond 3G mobile service provision, ser-vice adaptability, and network reconfigurabilitymanagement.

NANCY ALONISTIOTI ([email protected]) has B.Sc. and Ph.D.degrees in informatics and telecommunications (Universityof Athens). She worked for seven years at the Institute ofInformatics and Telecommunications of NCSR “Demokri-tos” in the areas of protocol and service design and test,mobile systems (UMTS), open architectures, CORBA, intelli-

gent networks, and software-defined radio systems andnetworks. She specializes in the formal specification andtesting of communication protocol and services, design ofobject-oriented SDL platforms, mobile applications, recon-figurability, adaptable services, design of test cases forconformance testing using TTCN, design of IN services,and CORBA-based architectures and services. She has par-ticipated in several national and European projects, (CTS,SS#7, ACTS RAINBOW, EURESCOM) and is technical man-ager of the IST-Mobivas project. She has worked as anexpert at the National Telecommunications Commissionand is currently project manager, senior researcher in theCNL (University of Athens) and member of the projectmanagement team of the Greek Academic Network,GUNET. Her current research includes mobile systems,software reconfigurable radio systems and networks,adaptability, service management and download, anddesign of open architectures and platforms.

It is the job of

the middleware to

abstract the

underlying network

services and

performance. Wehave developed a

mobile middleware

architecture to

support the concept

of seamless access,

breaking the

functionality down

into service support,

network abstraction,

and messaging

layers.

10.1.1.59.2500

Documents

Transcript of 10.1.1.59.2500