On the Vision of Complete Fixed-Mobile Convergence

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 28, NO. 16, AUGUST 15, 2010 2343

On the Vision of CompleteFixed-Mobile Convergence

Mohamed A. Ali, Georgios Ellinas, Hasan Erkan, Antonis Hadjiantonis, and Roger Dorsinville

(Invited Paper)

Abstract—This paper reviews and outlines the key features ofthe emerging next-generation fixed passive optical network (NG-PON)-based and fourth-generation (4G) mobile broadband accesstechnologies and how to leverage the advantages of both of theseaccess technologies to build a next-generation hybrid fiber-wireless(FiWi) network. Specifically, this paper presents an overview of themost recent research activities on these hybrid FiWi architectures,aiming to clarify the basic differences and distinguish between twoFiWi networking architectural models, namely, the overlay or in-dependent model and the truly integrated model. We then exploreand present a new direction to the design and implementation ofa simple and cost-effective all-packet-based converged fixed-mo-bile access networking solution that enables the true integration ofNG-PON and 4G mobile broadband access technologies into theenvisioned fixed-mobile platform. We briefly outline the generaland technical requirements to support a unified NG-PON long-term evolution ((LTE) radio access network (RAN) architecturethat conforms to both the typically centralized fixed PON and theemerging distributed 4G mobile LTE access standards. The imple-mentation methodology on how to efficiently and cost-effectivelyintegrate these two access technologies along with the main advan-tages gained from such an integrated architecture are outlined andpresented.

Index Terms—FiWi, optical-wireless, optical access networks,passive optical networks.

I. INTRODUCTION

P ASSIVE optical network (PON)-based fiber-to-the-curb/home (FTTC/FTTH) access networks are being de-

ployed around the world, using various architectures including

Manuscript received January 05, 2010; revised April 12, 2010; accepted April14, 2010. Date of publication May 27, 2010; date of current version August04, 2010. This work was supported in part by the Cyprus Research PromotionFoundation and the EU Structural Funds under Research Grant TPE/EPIKOI/0308(BIE)/06 and in part by the National Science Foundation, USA under GrantECCS-0901563.

M. Ali and R. Dorsinville are with the Department of Electrical Engineering,The City College of the City University of New York, New York, NY 10031USA (e-mail: [email protected]).

G. Ellinas is with the Department of Electrical and Computer Engineering,University of Cyprus, CY-1678, Nicosia, Cyprus.

H. Erkan was with the Department of Electrical Engineering, The City Col-lege of the City University of New York, New York, NY 10031 USA. He is nowwith the Department of Engineering, Maritime College, State University of NewYork, New York, NY 10018 USA.

A. Hadjiantonis is with the Department of Engineering and the Cyprus Aca-demic Research Institute, University of Nicosia, Nicosia 1700, Cyprus.

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/JLT.2010.2050861

single-channel time-division multiplexed (TDM) gigabit PON(GPON) and Ethernet PON (EPON) [1]–[6]. With the rapidgrowth in both wired and wireless data traffic and the sub-sequent needs of ever increasing bandwidth demand, there isan emerging interest in defining the next-generation of PON(NG-PON) access systems that are compatible with the currentGPON and EPON systems but with much higher bandwidth[7]–[9].

A PON connects a group of optical network units (ONUs)located at the subscriber premises to an optical line terminal(OLT) located at the service provider’s facility. To address thisissue, the full service access network (FSAN) community hascommissioned a study on defining possible smooth migrationscenarios from the current Gigabit-class PON systems towardthe NG-PONs, and their technical requirements, in order to se-lect the most suitable candidate system architectures that satisfythese requirements based on input from many operators [7]–[9].

The outcome of this study, which is eloquently summarized in[7]–[9], endorses two potential candidate system architecturesfor NG-PON. The first one is evolutionary growth architecture(termed NG-PON1), which supports coexistence with GPONon the same optical distribution network (ODN) and is viewedas a midterm upgrade. The second is a revolutionary disrup-tive architecture (termed NG-PON2) with no requirements interms of coexistence with GPON on the same ODN and is re-garded as a longer-term solution (assumed to be adopted afterNG-PON1). Several technical candidates have been proposed,including higher data rate TDM and dense wavelength-divisionmultiplexed (DWDM) systems. The FSAN community has con-cluded that extensive component Research and Development(R&D) is needed in order to make NG-PON2 an innovativeand cost-effective long-term solution. Another common salientfeature that has been underscored by the FSAN community isthat all NG-PONs including NG-PON1 and NG-PON2 must becapable of supporting multiple existing and emerging servicesacross multiple market segments, such as consumer, business,and mobile backhaul through their high quality of service (QoS)and high-bit-rate capabilities.

Concurrent with the upsurge of wireline PON-based broad-band access solutions, the growing demand for advanceddata-centric mobile multimedia services has accelerated thedevelopment and deployment of new wireless broadbandaccess technologies. These emerging technologies, includingthird-generation (3G) high-speed packet access (HSPA), 4Gmobile WiMAX, and cellular long-term evolution (LTE) are

0733-8724/$26.00 © 2010 IEEE

2344 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 28, NO. 16, AUGUST 15, 2010

capable of delivering speeds comparable to or better thancurrent fixed-line broadband access systems—up to 15–200Mb/s peak air throughput per user [10]–[15]. However, today’slegacy circuit switched T1/E1 wireline and microwave back-haul technologies cannot meet the capacity requirements ofwidespread HSPA, let alone 4G—so developing the backhaulnetwork is one of the primary challenges facing any mobileoperator looking to deploy a 4G network. Mobile backhaul,sometimes referred to as the radio access network (RAN), isutilized to backhaul traffic from individual base stations (BSs)to the radio network controller (RNC), which then connects tothe mobile operator’s core network (CN).

Mobile WiMAX and LTE are two different technologies thatwill eventually be used to achieve data speeds of up to 100Mb/s [10]–[15]. The crucial difference is that, unlike WiMAX,which requires a new network to be built, LTE runs on anevolution of the existing universal mobile telecommunicationsystem infrastructure already used by over 80% of mobilesubscribers globally. LTE is the latest RAN technology stan-dardized by the Third Generation Partnership Project (3GPP)to ensure the competitiveness of 3G for the next ten years andbeyond [10]–[15]. This is a part of a broader 3GPP systemcalled evolved packet system (EPS) that comprises a newall-IP mobile CN, the so-called evolved packet core (EPC) onthe core side and LTE on the access side. LTE consists of anew enhanced BS, called “evolved nodeB (eNB)” per 3GPPstandards. Specific EPC logical components are the mobilitymanagement entity (MME) in the control plane and the servinggateway (S-GW) and packet data network gateway (P-GW)in the bearer plane (see Section V shortly). In practice, bothgateways can be implemented as one physical network element[defined as access gateway (AGW)], depending on deploymentscenarios and vendor support.

By leveraging the advantages of both of these access tech-nologies combined on a truly integrated architecture platform,the long awaited vision of complete fixed-mobile convergence(FMC) could be materialized. By combining the practically un-limited capacity of optical fiber networks with the ubiquity andmobility of wireless networks, NG fiber-wireless (FiWi) net-works will enable the support of a wide range of emerging andunforeseen fixed-mobile applications and services independentof the access infrastructure [16]. Several hybrid FiWi sys-tems have been proposed including: 1) radio-over-fiber (RoF)systems [17]–[22]; 2) hybrid optical-wireless mesh network(WMN) architectures whose front end is essentially a multihopWMN with several wireless routers and a few gateways, whichconnect to the ONUs and, consequently, to the rest of theInternet through the OLT [23]–[27]; and 3) hybrid broadbandaccess architectures that attempt to integrate, or more preciselyto “overlay” PON and WiFi/WiMAX technologies [28]–[34].

In general, with the exception of the work reported in [35]that has outlined some brief but insightful qualitative analysison the potential of integrating EPON and WiMAX, these re-search efforts can be classified into three groups. The first grouphas mainly focused on investigating RoF transmission char-acteristics and modulation techniques, considering primarilyphysical layer-related performance metrics, e.g., power penalty,error vector magnitude (EVM), and bit-error-rate (BER) mea-

surements [17]–[22]. The second group’s efforts have mainlytargeted the performance of the front-end multihop WMN,including devising routing algorithms to optimally route datapackets across the WMN and finding the optimum locationsof the ONUs/gateway routers within the WMN [23]–[27]. Thethird group’s efforts [28]–[34] have mainly focused on charac-terizing the performance gain of the wireless segment when it iscollocated/overlaid with the fiber-based PON segment (ONUs)of the hybrid architecture. This is achieved via: 1) utilizing thehuge capacity offered via the fiber-based PON infrastructure tobackhaul mobile traffic; and/or 2) utilizing the centralized OLTto dynamically allocate upstream timeslots to the BSs usingtypical PON dynamic bandwidth allocation (DBA) schemes.For instance, the work presented in [34] utilizes first typicalPON DBA schemes to allocate timeslots (bandwidth) to theBSs. Then, each BS utilizes the assigned timeslot to provisiononly mobile users’ upstream traffic without considering at allthe fixed users attached to the PON. These DBA schemes aretailored to provision only one type of traffic (either fixed orwireless but not both) as they only take into account eitherthe fixed optical or mobile radio network resources (but notboth). Hence, the notion of a truly integrated DBA scheme thattakes into account the global network resources including bothfixed optical and mobile radio resources and, therefore, theprerequisite of truly integrated hybrid architecture has not beenaddressed.

Since the two access technologies utilized in most of theaforementioned referenced architectures are operating inde-pendently, the term “overlay architecture” is a more accuratecharacterization of most of the hybrid architectures reported inthe literature to date. To realize the full potential of combiningNG-PON and 4G mobile access technologies into the envi-sioned “truly integrated” fixed-mobile access infrastructure,at a minimum, two key techno-economic hurdles must beaddressed:

1) First, how to build a simple and cost-effective, mul-tiservice, converged all-packet-based fixed-mobile ac-cess networking transport infrastructure that conformsto both the typically centralized PON and the emerging4G distributed access standards (the first key enablingnetworking element). Equally important is then to as-sess and quantify the overall performance of such a con-verged infrastructure, specifically, in terms of its abilityto efficiently transport and support a mix of IP-based ad-vanced fixed and mobile multimedia traffic and servicesalong with the diverse QoS, rate, and reliability require-ments set by these services.

2) Second, how to truly integrate in terms of either software(unified control and management plane) or both hard-ware and software functionalities the two access tech-nologies of the hybrid architecture (the second key en-abling networking element).

It is important to emphasize that there is a radical distinction be-tween the general notion of FMC, which typically refers to con-vergence at the device level in which the same handset (dual-band mobile phones) has access to services through a fixed net-work in addition to a wireless network, and the notion pre-sented in this paper that targets a unified PON-RAN architec-

ALI et al.: ON THE VISION OF COMPLETE FIXED-MOBILE CONVERGENCE 2345

ture. FMC in this paper refers to convergence at the networklevel in which a customer can be offered mobile and fixed ser-vices seamlessly via a truly integrated fixed-mobile access net-working infrastructure.

To this end, this study explores a new direction to thedesign and implementation of a simple and cost-effectiveall-packet-based converged fixed-mobile access networkingsolution that enables the true integration of NG-PON and 4Gmobile broadband access technologies into the envisionedfixed-mobile platform. Specifically, this study proposes anddevises an all-IP/Ethernet fully distributed multiservice con-verged networking architecture that efficiently transports andsupports a wide range of existing and emerging fixed/mobileadvanced multimedia applications and services along with thediverse QoS, rate, and reliability requirements set by theseservices. To achieve this objective, we first build a compellingtechnoeconomic case for commercially deploying WDM-PONin the access arena as multiservice, all-packet-based convergedfixed-mobile optical-access networking transport architec-ture. We then introduce a novel and disruptive WDM-basedNG-PON architecture with native Ethernet, albeit with car-rier-class enhancements, as the converged packet-switchedaccess infrastructure (the first key enabling networking ele-ment listed earlier) along with several innovative networkingfeatures that collectively enable the support of a unified fullydistributed NG-PON-4G RAN access architecture. Becausethe salient features of the proposed NG-PON in this paper arealmost identical to those of the NG-PON2, specified by theFSAN community, the proposed architecture is also termedhere NG-PON2 to be consistent with the FSAN terminology.

Since LTE is emerging as the global standard for wirelesscarriers worldwide and has been positioned as the dominantNG mobile technology (e.g., several major U.S. wireless car-riers have already opted for LTE), in this work it will be se-lected as the 4G mobile access technology (wireless segmentof the proposed hybrid architecture). Though we have chosenEthernet-based PON and LTE as representative techniques forfixed PON and 4G mobile access technologies, the proposed ar-chitecture and related operation principles are also applicable toother PON and 4G access networks such as GPON and mobileWiMAX, respectively.

II. BUILDING THE CASE FOR DEPLOYING WDM-PON IN

THE ACCESS ARENA

Traditional WDM-PON systems allocate a separate pair ofdedicated upstream and downstream wavelength channels toeach subscriber, enabling the delivery of a symmetric 1 Gb/sor more of dedicated bandwidth per subscriber. In additionto its operational simplicity, WDM-PONs provide dedicatedpoint-to-point optical connectivity to each subscriber withbit rate and protocol transparencies, guaranteed QoS, andincreased security. Despite these numerous crucial advantages,WDM-PONs are still considered an expensive solution andhave not yet made any significant inroads into the currentaccess arena. This is mainly because the capacity offered byWDM-PON systems is still considered too high compared tothe access capacity needed to support typical current wire-line traffic demands and/or even higher future demands for

residential and commercial customers. As bandwidth demandincreases, however, the economics change. In terms of costper bit rate, WDM-PON is more efficient and economical. Aswe argue in this paper, utilizing the potential of WDM-PON’scapacity for both fixed and mobile backhaul applications andnetwork simplification, rather than just for blistering band-width to the home, is the key for the widespread acceptanceof WDM-PON in the access arena as a viable and dominantNG-PON access technology.

Recently, a combination of several key networking trendshave emerged that may collectively position WDM-PON as thedominant viable NG access transport infrastructure for boththe emerging data-centric fixed and mobile traffic and services.These include: 1) the inevitable migration from current legacyvoice-dominant TDM-based circuit-switched mobile backhaulinfrastructure to a new dynamic, all-packet-based mobile back-haul RAN architecture that is compatible with the emerginginherently packet-oriented NG mobile broadband access tech-nologies (e.g., HSPA and 4G mobile WiMAX and LTE); and 2)the inevitable trend toward all-IP/Ethernet transport protocolsand packet-switched networks. Ultimately, this will lead toan all-packet-based converged fixed-mobile optical transportnetwork from the core all the way out to the access network.Thus, the same packet-switched infrastructure can be utilizedto seamlessly transport both mobile and wireline business andresidential services.

Given the huge investments many fixed-line carriers aremaking or have already made in PON-based FTTH/FTTC ac-cess infrastructure, the combination of WDM-PON and nativeEthernet with a fiber-based access infrastructure is the mostpromising mix of technologies that ensures a cost-effectiveand future-proof converged fixed-mobile access transport in-frastructure. The economic advantage of utilizing the existingfiber-based PON access infrastructure with Ethernet function-ality is quite compelling compared to the choice of continuingto invest in legacy technology and/or the costly proposition ofbuilding up a new packet-based mobile backhaul infrastructure.With the planned migration to all Ethernet and IP transportprotocols and the introduction of Ethernet/IP interfaces inmobile base station and radio controller equipment, Ethernetis poised to become the dominant player in mobile telephonythat it has already become in the wireline voice and Internetdata networks. Furthermore, WDM-PON can efficiently andcost-effectively scale current mobile backhaul networks as wellas wireline TDM-PONs to meet the exponential increase inbackhaul capacity.

While the economics for commercially deployingWDM-PON in the access arena as a converged fixed-mobileoptical transport infrastructure are quiet compelling, however,several key outstanding technical challenges must be addressedfirst before WDM-PONs evolve as a truly viable optical accessnetworking technology that dominates future fixed-mobilebroadband access networking infrastructure. These are:

1) The converged WDM-PON access networking archi-tecture and topology must conform to both fixed PONand 4G mobile access standards. Exacerbating theproblem is that WDM-PONs are typically deployed astree topologies. However, a tree-based topology can


Fig. 1. Proposed converged WDM-based NG-PON-LTE RAN access architecture.

neither support a fully distributed access networkingarchitecture, nor an architecture with fully meshedaccess nodes (ONUs/BSs) in order to conform with theemerging 4G standards that require a new distributedLTE RAN architecture and further create a requirementto fully meshing the BSs [11]–[15].

2) To date, mainstream PON network control and manage-ment (NCM) operations including DBA and protectionschemes have been centralized, relying on a componentat the distant OLT to arbitrate upstream transmission andto detect and recover distribution and trunk fiber breaks.The centralized processes of bandwidth allocation aswell as the detection and subsequent recovery of bothtrunk and distribution fiber breaks at the distant OLT arelengthy and complex and require many changes at theOLT as well as each ONU. The challenge is how to rec-oncile the traditionally centralized PON’s NCM opera-tions and architecture with the typically distributed 4G’sNCM operations and architecture.

3) Typically, service resilience over previous generationsof PONs has not been a strong requirement from op-erators, with system field deployment close to nonexis-tent [7]. As a result, there is an obvious lack of simpleand cost-effective resilience capabilities, which guar-antee the reliable delivery of the emerging traffic andservices, specifically against failures in the distributionnetwork.

4) Inability to efficiently utilize limited available networkresources and to cope with the dynamic and burstytraffic patterns of the emerging data-centric fixed-mo-bile multimedia services. Since bandwidth is dedicatedon a point-to-point basis, there is no way to dynamicallymove capacity from a heavily loaded channel to anotherlightly loaded/idle channel.

Thus, the technical viability of WDM-PON as a converged ac-cess networking technology rests entirely on devising a simpleWDM-based NG-PON access networking paradigm that holis-tically addresses the previously listed challenges. Addressingthese challenges is precisely the main purpose of the proposedconverged NG-PON-LTE RAN access architecture (the first keyenabling networking element) that will be presented shortly.

III. PROPOSED CONVERGED WDM-BASED PON-LTE RANACCESS ARCHITECTURE

The proposed converged access architecture, as shown inFig. 1, is simply an innovative ring-based WDM-PON withnovel wireless-enabled OLT and ONUs, where each eNBis functionally integrated and/or collocated with the corre-sponding ONU and the EPC’s AGW is connected to the OLT.Since EPCs aggregate traffic from thousands of eNBs andmillions of subscribers, numerous WDM-PONs (OLTs) canbe attached to it (only two are shown in Fig. 1 for simplicity).Unlike a typical WDM metro ring network, where the feederfiber of a PON is replaced with a metro fiber ring that inter-connects the hub and access nodes, the proposed architectureinterconnects WDM access nodes (ONUs/eNBs) via a shortdistribution fiber ring in the local loop but allows them to sharethe feeder fiber for long-reach connectivity to the OLT/AGW.

As shown in Fig. 1, several access architectural models canbe used to support the collocation/integration of PON andLTE RAN infrastructures including independent and integratedmodels [34]. For the independent model, the PON and 4Gsystems are operated independently by considering an LTE BS(eNB) a generic user attached to an ONU and/or collocated withit. The RAN architecture is assumed to have its own operation,control, and management independent of those for the PONarchitecture. The ONU and eNB can be interconnected, as longas they support a common standard interface (e.g., Ethernet).


Fig. 2. Proposed stand-alone ring-based WDM-PON architecture.

Thus, the OLT, AGW, ONUs, and eNBs are all assumed tosupport a common standard interface (e.g., 802.3(a)h Ethernetinterface). Under the integrated model, an ONU and a cellularBS can be integrated into a single module in terms of eithersoftware or both software and hardware functionalities. Be-cause the integrated model, rather than the overlay model, isthe main theme of the proposed architecture, the details of thefunctional integration of both modules will be presented inSection V shortly.

For simplicity, throughout the remaining of this paper, the no-tations used for typical PONs will also be used for the proposedconverged architecture. Thus, the word “ONU” is often used, butshould also be taken to mean “ONU/eNB”. Similarly, the word“traffic” should also be taken to mean “fixed and mobile traffic”for all downstream, upstream, and LAN traffic. The followingsections present an overview of the stand-alone WDM-basedNG-PON architecture along with a brief summary of each of itssalient feature that collectively enable the architecture to sup-port the requirements listed in Section II mentioned earlier.

A. Overview of the Stand-Alone WDM-Based NG-PONArchitecture (NG-PON2)

Fig. 2 illustrates the proposed stand-alone ring-basedWDM-PON architecture. An OLT is connected to N ONUs(this work assumes ONUs) via a 10–20 km trunkfeeder fiber, a passive 3-port optical circulator, and a smallfiber ring [36]. To cover the same local access area as in thetree-based architecture, the small ring at the end of the trunkis assumed to have a 1–5 km diameter. The ONUs are joinedtogether by point-to-point links in a closed loop around theaccess ring. The links are unidirectional: both downstreamand upstream signals (combined signal) are transmitted inone direction only. Each ONU is assigned a single-dedicatedwavelength for both downstream and upstream transmissions.Direct intercommunication among ONUs is achieved via anadditional local control/LAN wavelength channel, , whichis terminated, regenerated, and retransmitted at each ONU.

The OLT houses an array of N fixed transmitters (Tx) and an-other array of fixed receivers (Rx), a passive 3-port optical

circulator, and a low-cost thin-film-based WDM multiplexer/de-multiplexer with channel-dependent insertion loss between 0.3and 5 dB. Each Tx/Rx pair corresponds to one ONU and utilizesthe same wavelength for transmitting and receiving downstreamand upstream traffic, respectively. The extra receiver lo-cated at the OLT is used to detect the local control/LAN channel.Each ONU has a Tx/Rx pair, which is matched to the corre-sponding pair at the OLT and another Tx/Rx pair for transmit-ting and receiving the local LAN channel, . In addition,each ONU houses a low-cost 6-port thin-film filter-based fixedoptical add--drop multiplexer (OADM), where two wavelengths(corresponding downstream/upstream and LAN wavelengths)are dropped and added at each node.

The downstream WDM signal is coupled to the ring via port3 of the optical circulator. After recombining it with the recircu-lated LAN signal via a 2 1 WDM multiplexer (placed on thering directly after the optical circulator), the combined signalthen circulates around the ring ( through ) in adrop/add and go-through fashion. For instance, at the first ONU,the dedicated downstream wavelength channel along with therecirculated control/LAN channel are dropped and processed;then, the dedicated upstream signal along with the regener-ated control/LAN channel are added. Finally, at the last node

, wavelengths and are dropped/added. Thus,the downstream WDM signal is terminated at the last node.

The combined upstream WDM and LAN signal emergingfrom the last ONU at the end of the ring is split into two compo-nents via a (10:90) 1 2 passive splitter placed on the ring di-rectly after the last ONU. The first component (90%) is directedtoward the OLT via circulator ports 1 and 2, while the secondcomponent (10%) passes first through a band rejection filter thatterminates the upstream WDM signal. The second component,i.e., the LAN signal emerging from the band rejection filter, isallowed to recirculate around the ring after recombining withthe downstream signal (originating from the OLT) via the 2 1CWDM combiner. The first component, i.e., the combined LANand upstream WDM signal, is received and processed by thearray of fixed optical receivers (housed at the OLT).Specifically, each of the N upstream optical receivers detectsthe corresponding upstream signal and recovers the MAN/WAN


traffic, while the LAN optical receiver processes the controlmessages and, as will be shown later on, may discard or processthe LAN traffic, depending on whether it is pure LAN traffic ora mix of both upstream and LAN traffic. Assuming that the pro-posed architecture can support up to 16 ONUs, downstream/up-stream wavelengths as well as the LAN wavelength are eitherspaced 200-GHz apart in the 1530–1565 nm C-band or can beallocated over the 1270–1610 nm CWDM spectrum (can offerup to 18 available channels) with 20-nm spacing, as defined inITU TG.694.2. In the latter case, the overall system cost can bereduced via the utilization of low-cost commercially availableCWDM components. To scale beyond 16 ONUs, the numberof downstream/upstream wavelength channels can be doubledor quadrupled by reducing the channel spacing in the C-banddown to 100- or 50-GHz, respectively.

IV. KEY ENABLING NETWORKING FEATURES

This section outlines the key four salient networking fea-tures that enable the support of a unified fully distributedNG-PON-LTE RAN access architecture.

A. Fully Distributed Unified Control Plane

This study utilizes the control and management messagesdefined by the IEEE 802.3(a)h multipoint control protocol(MPCP) standard [1] that facilitate the exchange of control andmanagement information between the ONUs/eNBs and OLT.The protocol relies on two Ethernet control messages, GATE(form OLT to ONUs/eNBs) and REPORT (from ONUs/eNBsto OLT and between ONUs/eNBs) messages in its regular op-eration. Direct communication among ONUs/eNBs is achievedvia a separate dedicated LAN/control wavelength channel,which is terminated, processed, regenerated, and retransmittedat each ONU/eNB. Since control messages are processed andretransmitted at each node, the ONUs can directly communi-cate their upstream/LAN queue status and exchange signalingand control information with one another in a fully distributedfashion.

Likewise, eNBs can also directly communicate the statusof their queues and radio resources and exchange signalingand control messages along with their own local mobile LANtraffic with one another. The control plane utilized among theONUs/eNBs can thus support a distributed PON-RAN archi-tecture, where each access node (ONU/eNB) deployed aroundthe ring has now a truly physical connectivity and is, thus,capable of directly communicating with all other access nodes,in conformity with LTE standards. The control plane utilizes atime-division-multiple access (TDMA) arbitration scheme toimplement fully distributed QoS aware DBA and packet sched-uling algorithms in which the OLT/AGW is excluded from theprovisioning process. Thus, supported by the distributed controlplane, most of the typical radio control functions includingradio resource management, handover control, admission con-trol, etc., can be independently implemented at each eNB ina distributed approach without resorting to a central controlentity (e.g., RNC/AGW), in conformity with LTE standards.

Likewise, most of the typical wireline control functionalitiesincluding DBA, queue management, packet scheduling, restora-tion algorithms and schemes, etc., can be independently im-plemented at each ONU in a distributed approach without re-sorting to a central control entity (e.g., OLT). These function-alities are typically implemented at the distant OLT/RNC intoday’s stand-alone centralized PON/RAN systems. Please notethat, under the overlay model, the distributed control plane en-ables each eNB attached to the ring to independently provisiononly mobile traffic (upstream and LAN traffic). Likewise, it alsoenables each ONU to independently provision only fixed traffic(but not both).

Under the integrated model, however, as will be shown inSection V, in addition to the ONU’s and eNB’s control mod-ules (typically supported by the overlay model), the inclusionof an additional common third control module that interfacesto both the PON and eNB control modules enables the sup-port of a unified control and management plane. ONU’s moduleinterfaces with the PON section and runs the PON protocols;eNB’s module interfaces with LTE section and runs the LTEprotocols. The common module interfaces to both the PON andeNB sections, manages and coordinates joint optical-radio re-sources, and executes the integrated DBA and packet schedulingalgorithms. Thus, the control plane now is both distributed andunified. Supported by such a unified distributed control plane,the processes of bandwidth allocation, packet scheduling, andqueue management for both the PON and RAN systems are in-tegrated and operate in a distributed manner. Specifically, eachaccess node (ONU/eNB) can independently provision both fixedand mobile upstream traffic (ONUs-OLT, eNBs-OLT-EPC, andUEs-eNBs) as well as LAN traffic.

B. Dynamic Bandwidth Assignment and Sharing of a Mix ofDownstream, Upstream, and LAN Traffic

In this paper, as will be detailed shortly, LAN data may be acombination of native local LAN traffic, transient downstream(TDS) traffic, and transient upstream (TUS) traffic. TDS trafficis defined here as only low-priority best-effort downstreamtraffic (the fraction of bursty downstream fixed-mobile trafficthat may exceed the user’s dedicated downstream wavelengthchannel rate) destined to a given access node but ter-minated at a different transient access node . This isbecause TDS traffic is always transported via a nondedicateddownstream wavelength channel other than its own dedi-cated downstream wavelength channel, , which is typicallypre-assigned to transport native ’s downstream traffic.Once TDS traffic is terminated at , it is handled astransient LAN traffic whose new source is and finaldestination is . TDS traffic along with ’s localLAN traffic are first buffered at the corresponding ’sLAN queue, and are then retransmitted over the LAN/controlwavelength channel around the ring to their final destinations,within the same granted ’s LAN timeslot.

Similarly, TUS traffic is defined here as only low-prioritybest-effort upstream traffic (the fraction of bursty upstreamfixed-mobile traffic that may exceed the user’s dedicated up-stream wavelength channel rate) originating from a source node


and destined to OLT, but transported via a wavelengthchannel other than its own dedicated upstream wavelengthchannel , namely, the LAN/control channel, . SinceLAN transmission is based on a TDMA scheme, inter-ONUtraffic including local LAN data and control messages as wellas TDS and TUS traffic, are all transmitted within the samegranted timeslot.

Local mobile LAN traffic is defined here as bidirectionaltraffic sourced from a mobile user that is served by a givenBS (access node) attached to the ring and destined for anothermobile user that is served by another BS (different accessnode), which is also attached to the same ring (same PON do-main). This traffic is directly routed on the ring from the sourceeNB directly to the destination eNB and vice versa as LANtraffic, without the direct participation of either the OLT or theEPC. This local traffic represents bidirectional upstream dataexchange (including VoIP, video, and data sessions) betweenany two mobile users served by two different eNBs that areattached to the same ring. This is significant as the volumeof voice calls and/or multimedia data exchange between localmobile users is substantial.

Since the LAN/control channel is shared among all ONUs/eNBs, a distributed DBA scheme is required to efficiently andfairly provision local LAN traffic among ONUs/eNBs. Specifi-cally, since the control information only occupies a fixed smallfraction (5%–10%) of the control channel cycle, the DBA algo-rithm can dynamically allocate the remaining cycle bandwidthto local LAN traffic. The distributed LAN DBA algorithm iscycle based, where a cycle is defined as the time thatelapses between two executions of the scheduling algorithm. Acycle has a variable length size confined within certain lowerand upper bounds, which we denote as and .

Each access node maintains a database about the states of itsqueue and the state of every other ONU/BS’s queue on the ring.This information is updated each cycle whenever the ONU/eNBreceives new REPORT messages from all other ONUs/eNBs.During each cycle, the access nodes sequentially transmit theirREPORT messages along with LAN data in an ascending orderwithin their granted LAN timeslots around the ring from onenode to the next, where each REPORT message is finally re-moved by the source ONU after making one trip around the ring.The REPORT message typically contains the desired size of thenext LAN timeslot based on the current ONU’s LAN buffer oc-cupancy. Note that the REPORT message contains the aggregatebandwidth of both fixed and mobile LAN data buffered at eachONU’s/eNB’s queue (requested size of next LAN timeslot).

An identical LAN DBA module, which resides at each ONU/eNB, uses the REPORT messages during each cycle to calculatea new LAN timeslot assignment for each ONU. ONUs sequen-tially and independently run instances of the same LAN DBAalgorithm outputting identical bandwidth allocation results eachcycle [4]–[6]. The execution of the algorithm at each ONU startsimmediately following the collection of all REPORT messages.Thus, all ONUs must execute the DBA algorithm prior to theexpiration of the current cycle so that bandwidth allocationsscheduled for the next cycle are guaranteed to be ready by theend of the current cycle. An execution of the DBA algorithm

produces a unique and identical set of ONU assignments. It iscritical that the algorithm produces a unique outcome for any ar-bitrary set of inputs. Once the algorithm is executed, the ONUssequentially and orderly transmit their data without any colli-sions, eliminating the OLT’s centralized task of processing re-quests and generating grants for LAN bandwidth allocations.

The distributed LAN DBA algorithm utilized here is typi-cally a single-channel TDM-PON DBA algorithm [2]–[6], butmodified here to account for both TDS and TUS traffic. Thereader is referred to [5], [6] for further details regarding thesingle-channel distributed LAN DBA algorithm. The modifiedalgorithm provisions native LAN data as well as TDS and TUStraffic, but handles all these different types of data as just oneclass of data (LAN data) and does not distinguish between them.Thus, when an ONU reports its LAN queue status to the otherONUs as well as to the OLT, it must periodically report the ag-gregate bandwidth of all data traffic buffered atits LAN queue without any distinction, including its own localLAN traffic as well as TDS and TUS traffic (if there is any). TheDBA module housed at each ONU/eNB uses this informationduring each cycle to calculate one new LAN timeslot assign-ment for each ONU/eNB.

Note that maintaining accurate time synchronization betweenthe ONUs is essential for the appropriate operation of the dis-tributed DBA algorithm. In general, this is always the case, as allONUs are synchronized to a common reference clock extractedfrom the OLT downstream traffic. Clocking information, in theform of a synchronization marker, is encapsulated at the begin-ning of each downstream wavelength frame cycle. The synchro-nization marker is a one-byte code that is transmitted every 2 msto synchronize all ONUs with the OLT [1]. The TDM controllerat each ONU, in conjunction with timing information from theOLT that must takes into account the ranging problem (the dis-tance, and hence propagation delay, between the OLT and dif-ferent ONUs on the ring is not the same), controls the LANtransmission of the variable-length packets within the dedicatedtime slots.

Using this scheme we can now effectively separate traffic ofdifferent service classes (or different IP flows), by assigning theentire dedicated upstream/downstream wavelength channels tohigh-priority traffic (e.g., VoIP and real-time streaming traffic)while assigning the remaining cycle bandwidth of the con-trol/LAN channel to local LAN traffic as well as low-prioritybest-effort traffic, i.e., TDS and TUS traffic. If more than onecongested downstream/upstream channels are contending forthe remaining limited capacity of an available lightly loadeddownstream/LAN channel, the scheduler at the OLT/ONUmust apply some traffic-engineered selection criteria includingfairness along with the user’s service level agreement to selectone from amongst the congested downstream/upstream chan-nels. As per the LAN traffic, local fixed and mobile traffic havealways higher priority than both TDS and TUS traffic.

It is important to emphasize that a distributed PON-RANarchitecture is a prerequisite for implementing an efficientdynamic wavelength/timeslot sharing strategy for upstream,downstream and LAN traffic, since each ONU/eNB mustindependently provision its own TDS, TUS, and LAN traffic.


Due to the significance of the LAN channel, its line rate, atthe initial stage of deploying the access network, can be set inaccordance with the projected total network traffic load (e.g., 1,2.5, or 10 Gb/s).

C. Downstream Wavelength Assignment and Sharing

The OLT houses 16 downstream queues; each queueis assigned to a specific and is connected to adedicated downstream wavelength . Each ONU houses threequeues, one queue , is assigned to a dedicated upstreamwavelength, another queue, , is assigned to the ter-minated downstream wavelength , while the third queue

is assigned to the LAN/control traffic. The processof dynamically assigning/sharing downstream wavelengths isimplemented jointly at both the OLT and ONUs as follows:if a dedicated downstream wavelength channel with trafficdestined to is overloaded (i.e., incoming bursty trafficflows may exceed the dedicated channel rate for some interval,so that its corresponding is congested), the followingsteps are executed.

1) The scheduler at the OLT searches for another under-utilized/idle downstream wavelength channel , i.e., achannel whose corresponding queue has someavailable space that can accommodate one or more of

’s excess flows [termed here as transit downstream(TDS) flows/traffic]. This queue will be referred to asan available queue.

2) If the search is successful, the available channel, , isselected if and only if its corresponding LAN queue at

, is also available.3) The scheduler redirects one, some, or all of ’s excess

flow(s) to (only low-priority best-effort traffic),where it is then transmitted, along with ’s na-tive downstream traffic, to over its dedicatedwavelength channel . Such a channel that can ac-commodate and transport, in addition to its own localdownstream traffic, other wavelength channel’s down-stream traffic, will be referred here as an “acceptorwavelength channel (AWC)” and its correspondingqueue as an “available ready queue.”

4) The -downstream optical receiver housed atterminates all ’s downstream traffic including bothnative downstream traffic destined to and TDStraffic destined to examines the destination MACaddress of each detected Ethernet frame, and then per-forms the following two functions: 1) native downstreamtraffic that matches ’s MAC address is copiedand delivered to the end-users; 2) TDS traffic destinedto (whose MAC address does not match that of

) is redirected to ’s LAN queue and then re-transmitted, along with ’s own local LAN traffic,as LAN traffic around the ring over to its final des-tination, within the proper designated LAN timeslot of

.5) TDS flows that are still buffered at the OLT’s queues are

returned back to their original queues once these queuesare available.

D. Upstream Traffic Flows Rerouting and Sharing

Analogous to traditional WDM-PONs, each ONU in the pro-posed architecture is assigned a dedicated wavelength for up-stream transmission. However, if the incoming user’s burstytraffic flows exceed its dedicated upstream wavelength channelrate for some interval, the corresponding upstream queue

becomes congested. In this case, the flow scheduler atmay redirect one or more of the user’s excess upstream

service flows to the local LAN queue (these flows are now TUSflows) provided that this LAN queue has some avail-able space that can accommodate one or more of ’s excessflows.

Once these TUS flows are buffered at , they arehandled as transient LAN traffic and are transmitted, alongwith ’s local LAN traffic as well as TDS flows (if thereis any) via the LAN/control channel to their final destinations,within the same ’s granted LAN timeslot. Since the LANchannel is split into two components at the end of the ring (thefirst component is destined to the OLT/RNC, while the secondone recirculates around the ring), the LAN optical receiverhoused at the OLT detects the first component of the LANchannel, recovers and processes TUS flows as well as controlmessages, and discards native LAN and TDS traffic (if there isany). On the other hand, the source ONU must remove both itsown TUS flows and REPORT message from the recirculatingsecond component of the LAN channel. TUS flows are returnedback to their original upstream queues once they are available.

The reader should note that the implementations of theresource allocation schemes, distributed DBA and schedulingalgorithms, dynamic threshold used to determine whether aqueue is congested or available, and traffic-engineered routingof end-user’s fixed/mobile traffic flows, at both the OLT andthe access nodes (OUNs/eNBs), needed to efficiently supportsdynamic allocation of downstream wavelength channels aswell as sharing downstream, upstream, and LAN traffic amongPON-RAN fixed-mobile end-users is beyond the scope of thispaper and the reader is referred to [36] where some of theseschemes have been already developed but only for fixed PONusers. The simulation results presented in [36] demonstratesome of the key system performance metrics including networkutilization and further verify the effectiveness and robustnessof the proposed schemes.

V. FUNCTIONAL INTEGRATION OF THE TWO

ACCESS TECHNOLOGIES

The EPS represents a migration from the traditional hierar-chical system architecture to a flattened architecture that min-imizes the number of hops and distributes the processing loadacross the network. To illustrate some of the key technical chal-lenges associated with such functional integration, it is first im-portant to understand the novel and radical changes associatedwith the evolving LTE RAN architecture. First, the 3G-RNC iseliminated from the data path and its typical functions are in-corporated into the eNB, including all radio control functionssuch as radio resource management, handover control, admis-sion control, etc. Thus, the distributed nature of the LTE RAN


Fig. 3. (a) Architecture of the ONU-eNB, (b) functional modules hardware layout.

architecture calls for new radio control algorithms and proce-dures that operate in a distributed manner, including distributedhandover schemes as well.

Second, with RNC functionality distributed to the eNBs, LTEcreates a requirement for fully meshing the eNBs—some 10 000to 40 000 for a mobile operator running a network in a “ma-ture” market. The cause is call handoff (HO). A call originatingon the “anchor” eNB that subsequently travels across the net-work—moving from eNB to eNB requires that each eNB uti-lized will need to communicate with both the anchor eNB (e.g.,for billing purposes) and the previous and next eNB for call HO[37]. Even though IP-based virtual private networks (VPNs) canbe considered good candidates for this type of meshing, how-ever, to support a mesh of this size, some degree of hierarchyis likely to be required. A hierarchical IP VPN may providean ideal migration path for LTE and the all-IP RAN. To realizethe full potential of LTE, mobile operators must deploy such akind of networking architecture in the future; however, the com-plexity as well as the highly prohibitive cost of such a large-scaledeployment might make its implementation rather a distant fu-ture. The implications of these radical changes are significantas they directly impact the proposed converged architecture, theleast of which is that the proposed architecture must complywith these sweeping requirements as well.

The proposed converged architecture eloquently complieswith both of these radical changes via the purposely selectedsimple ring topology, which enables direct intercommunica-tion/connectivity among the ONUs/eNBs, allowing for thesupport of a distributed PON-RAN access architecture aswell as meets the stringent requirement to fully meshing theONUs/eNBs. Thus, the proposed ring-based architecture mayprovide a simple and cost-effective solution at the initial phaseof deploying LTE. Given the simplicity, inherent resilience,and ubiquity of the ring architectures, a full-scale deploymentof numerous interconnected local access WDM rings (as theone proposed here) with one or more access node(s) acting asgateway(s) nodes (IP routers) that interconnect one or morelocal WDM rings, may also provide a simple and cost-effectivescenario to fully meshing such a large number of eNBs, whichwill be needed in a mature LTE market.

A. Integrated Model

Under the integrated model, an ONU and a cellular LTE’sBS (eNB) can be integrated into a single module either in termsof software or both software and hardware functionalities. Thefollowing are the main technical requirements needed to supportthe functional integration of the PON and 4G LTE access infra-structure: 1) each dedicated wavelength channel (assigned to agiven ONU/eNB) is treated as single-channel TDM Ethernet-based PON (EPON) system and is further assumed to complywith all the IEEE 802.3(a)h EPON standards; 2) the OLT, S-GW,ONUs, and eNBs are all assumed to support a common standardinterface (e.g., 802.3(a)h Ethernet interface); 3) each ONU is as-sumed to have two different Ethernet port ranges, the first portrange will support wired users, while the second port range willsupport mobile users. The port ranges will be used by the ONUsto identify and differentiate between mobile versus fixed users;4) in Fig. 1, all the intermediate nodes (e.g., OLT, ONU, eNB)are assumed to be equipped with an IP access router to forwardIP packets.

Fig. 3(b) illustrates the three main control modules of thefunctionally integrated ONU-eNB access node, namely, ONU’scontrol module, eNB’s control module, and the common con-trol module, where each module can be a single CPU in hard-ware [35]. ONU’s module interfaces with the PON section andruns the PON protocols; eNB’s module interfaces with LTE sec-tion and runs the LTE protocols. The common module inter-faces to both the PON and eNB sections, manages and coordi-nates joint optical-radio resources, and executes the integratedDBA and packet scheduling algorithms. ONU and eNB mod-ules report their queue statuses and bandwidth request details tothe common module; the latter utilizes this received informationto make decisions, and to optimally allocate upstream/LAN re-sources to the ONUs and eNBs.

The functional modules for provisioning upstream trafficcorresponding to the three modules in Fig. 3(b) are shown inFig. 3(a). Specifically, the ONU’s control module that interfaceswith the PON section includes the functional components ofPON packet scheduler, priority queues management, and PONpacket classifier. Similarly, the LTE’s module that interfaces tothe LTE section includes the functional components of two LTE


mapping modules [one to map user equipment’s (UE’s) radiobearers to mobility tunnels], eNB packet classifier, and LTEupstream scheduler. Finally, the third module at the bottom ofFig. 3(a) corresponds to the ONU-eNB common coordinatorcontroller, which comprises the functional components re-quired to map QoS between PON and LTE and performs globaladmission and congestion control as well as integrated DBAand resource allocation and sharing protocols and algorithms.

B. Unified Distributed Control and Management Plane

It is important to emphasize that the distributed control planedescribed previously that manages the intercommunicationamong the ONUs/eNBs can enable each and every ONU/eNBattached to the ring, even under the overlay model, to indepen-dently provision either fixed or mobile upstream/LAN traffic,but not both (since the third common control module is absent).However, under the integrated model, the inclusion of the thirdcommon control module enables the support of a unified controland management plane that manages and controls the overallfixed optical and mobile radio network resources. Thus, thecontrol plane now is both distributed and unified. Supported bysuch a unified distributed control plane, the processes of band-width allocation, packet scheduling, and queue managementfor both the PON and RAN systems are integrated and operatein a distributed manner. Thus, the integrated model provides thebest overall system performance in terms of cost-effectiveness,bandwidth utilization, and support for better QoS.

This is because the integrated control module has globalinformation about the entire fixed/mobile network status in-cluding the aggregate bandwidth requirements of both wiredand mobile users. Hence, the processes of bandwidth allocationand packet scheduling as well as prioritizing different classof services (for either wired or mobile users) are optimizedthroughout the network system. Supported by the distributedcontrol plane, most of the typical wireline and wireless controlplane functionalities such as DBA, queue management andpacket scheduling, HO control, radio resource management,admission control, etc., typically implemented in today’sstand-alone centralized OLT/RNC, are now incorporatedinto the functionally integrated access nodes (ONUs/eNBs).These are independently implemented at each access node(ONU-eNB) in a distributed approach without resorting to acentral control entity (e.g., OLT/RNC), in conformity with LTEstandards.

C. QoS Support and Mapping

The QoS model of EPS, which was standardized in 3GPPrelease 8, is based on the logical concept of an “EPS bearer”[10]–[15]. The term “bearer” refers to a logical IP transmissionpath between the UE and the EPC with specific QoS parame-ters (capacity, delay, packet loss error rate, etc.). Each bearer isassigned one and only one QoS class identifier (QCI) by the net-work and is composed of a radio bearer and a mobility tunnel.The QCI is a scalar that is used within the access network toidentify the QoS characteristics that the EPC is expected to pro-vide for the IP service data flows (SDFs). This scalar (bearer ID)is used by routers to access node-specific parameters that con-trol packet forwarding treatment (e.g., scheduling policy, admis-

sion thresholds, link layer configurations, queue managementpolicy, etc.), which are specified and preconfigured by the oper-ator. An EPS bearer uniquely identifies packet flows that receivethe same packet forwarding treatment between the UE and EPC.Thus, the aggregated IP flows constituting a bearer are carriedfrom the UE over the radio interface to the eNB, from the eNBto the S-GW, and then onward to the P-GW as a single-logicalbearer with the same level of QoS (or packet forwarding treat-ment). Services with IP flows requiring a different packet for-warding treatment would therefore require more than one EPSbearer.

Furthermore, an IP flow is defined by a five-tuple (the sourceand destination IP addresses, source and destination port num-bers, and protocol ID), which is used by the packet filter to iden-tify different IP flows. Downlink (DL) IP flows are identifiedby DL packet filters located at the P-GW, while uplink (UL)IP flows are identified via UL packet filters located at the UE.Thus, the UE/P-GW performs UL/DL packet filtering to map theoutgoing/incoming IP flows onto the appropriate bearer (bearerbinding). There are two types of bearers: guaranteed bit-rate(GBR) and nonguaranteed bit-rate (non-GBR) bearers. A GBRbearer has a GBR and maximum bit-rate (MBR), while morethan one non-GBR bearer belonging to the same UE shares anaggregate maximum bit rate (AMBR). Non-GBR bearers cansuffer packet loss under congestion, while GBR bearers are im-mune to such losses (via admission control functions that resideat the eNB and P-GW). A bearer can also be classified as ei-ther a default or a dedicated bearer. The default bearer is setup when the UE attaches to the network to provide the basicconnectivity. The 3GPP specifications mandate that the defaultbearer is a non-GBR bearer. The dedicated bearer can be eithera GBR or a non-GBR bearer.

For a given bearer, QoS characteristic is completely definedby two parameters: QCI (bearer ID) and allocation and reten-tion priority (ARP) that specifies the control plane treatment thatthe bearers receive. ARP does not have any impact on packetforwarding behavior but is used to decide whether a bearer es-tablishment/modification request can be accepted or rejected.The 3GPP specifications define eight standardized QCIs, eachwith its corresponding standardized characteristics, includingbearer type (GBR versus non-GBR), priority, packet delay, andpacket-error-loss rate. To allow for traffic separation in the trans-port network (IP cloud connecting the eNBs to the EPC), P-GWand eNB map each QCI onto a corresponding diffserv codepoint (DSCP) in order to translate a bearer-based QoS (QCI) totransport-based QoS (DSCP) [12]. Using this mapping function,packets on a bearer associated with a specific QCI are markedwith a specific DSCP for forwarding in the transport network.The QCI to DSCP mapping is performed based on operator poli-cies, which are configured into the network nodes. P-GW per-forms the mapping for DL packets while eNB performs it forUL packets.

As can be seen from the eNB module, as shown in Fig. 3(a),the UE uses the packet filters to classify IP packets to authorizedIP SDFs. This process is referred to as SDF detection. The UEthen performs the binding of the detected UL IP SDFs to the ap-propriate bearers. Once the UE’s radio bearers are terminated atthe eNB, they are mapped into the appropriate mobility tunnels


based on their bearer IDs. The eNB’s packet classifier then mapstheir constituent IP flows into their appropriate priority queuesbased on the bearer IDs attached to the IP packets, which is thebasic enabler for traffic separation. Finally, to allow for trafficseparation in the transport network, the eNB maps each OCI(bearer-ID) onto the corresponding DSCP value.

On the other hand, EPON technology does not support thistype of bearer-based connection. Rather, bandwidth requestsare queue oriented; an aggregate bandwidth is allocated to eachONU, and then the latter makes a local decision to allocate thegranted bandwidth and schedules packet transmission for up toeight different priority queues in the ONU. Both EPON andLTE classify data traffic in a differentiated services mode. How-ever, EPON supports only enhanced QoS through prioritizationwhere packets are classified, stored in different priority queuesand, then, scheduled for service according to their priority. Onthe other hand, LTE supports guaranteed QoS through logicalbearer reservation where each router/node on the RAN/EPC isconfigured to forward the packets of different IP flows based ontheir bearer-IDs (QCIs) in which resources are reserved (queuespace, queuing management strategy, scheduling strategy) ac-cordingly.

To achieve a truly integrated model, an effective mappingmechanism is required between EPON priority queues andQCI/bearer-based LTE IP flows. Specifically, the mappinghas to identify which LTE IP flow should be stored in whichEPON priority queue for equivalent QoS. EPON has up to eightdifferent priority queues in each ONU, while LTE defines eightstandardized QCIs that classify data traffic into eight differentclasses of service, ranging from real-time gaming to the lowestpriority best-effort TCP bulk data. This theoretically facilitatesa one-to-one mapping from eNB’s eight priority queues toONU’s eight priority queues (e.g., packets of highest/lowesteNB’s priority queue are mapped onto highest/lowest ONU’spriority queue) and vice versa in both upstream and down-stream directions. However, devising an efficient viablemapping strategy that enables a unified QoS model for bothwired and wireless services requires the implementation of thefollowing critical functions.

1) Since the bearers are not visible to the ONUs/OLTs,each and every ONU/OLT must be directly configured(semistatically) with all eight LTE’s standardized QCIs(QoS characteristics) or more precisely with the cor-responding DSCP values (QCI to DSCP mapping isperformed based on operator policies). This configura-tion enables each ONU/OLT to forward the packets ofdifferent UL/DL IP flows based on their DSCP valuessuch that the packets-forwarding treatment received bythese flows at the ONU/OLT is identical to that receivedat the eNB/P-GW. This is achieved by ensuring that thequeue management schemes and scheduling algorithmsimplemented at the ONU/OLT are identical to those im-plemented at the eNB/P-GW.

2) The PON’s packet scheduler at the ONU/OLT must applythe same packet forwarding treatment for both wired andwireless upstream/downstream traffic for each and everyconfigured QCI/DSCP value that is associated with a givenIP flow. This further enhances the typical PON’s prioriti-

zation-based QoS support for wired users as well as sim-plifies the implementations of queue management schemesand scheduling algorithms at the ONUs and OLTs.

3) The typical PON’s cycle-based approach for DBA andQoS support must be drastically modified at both theONUs and OLTs. Typical centralized PON architecturessupport differentiated upstream QoS via two indepen-dent scheduling mechanisms: a) scheduling at the OLT(inter-ONU scheduling/DBA); the OLT arbitrates up-stream transmission by periodically (each cycle, withthe maximum duration of the cycle around 2 ms) allo-cating an appropriate timeslot (bandwidth) to each ONU;b) scheduling at the ONU (intra-ONU scheduling); theONU makes local decisions to allocate the granted band-width and schedules packet transmission for each priorityqueue. However, none of these scheduling mechanismscan guarantee bandwidth for real-time IP flows becausethe bandwidth allocated by the OLT to one ONU canonly be guaranteed for a significantly short time (e.g., afraction of one cycle) and may vary from one cycle toanother according to the load at other ONUs. Thus, eachONU is required to reserve bandwidth for its real-time IPflows for the whole duration of the flow (and not on a percycle basis) in order to satisfy their QoS requirements asspecified by the attached DSCP value.

4) In addition to bandwidth allocation and service differen-tiation, a global admission and congestion control (AC)mechanism for both wired and wireless traffic is requiredthat makes decisions on whether or not to admit/block anew wired/wireless real-time IP flow based on its require-ments and the upstream channel usage condition. Ideally,this AC module should be housed at the common controlmodule (see Fig. 3) since the critical information neededby the AC module to make appropriate admission/denialdecisions (e.g., available fixed optical and mobile radio re-sources as well as both available wired (ONUS-OLT) andwireless (UE-eNBs) UL channel capacities) is always dy-namically available to the common control module. For in-stance, when the congestion bottleneck is at the backhauland not at the radio interface, the common control modulecan block the admission of any new mobile user’s trafficuntil congestion subsides.

The combination of a distributed PON-RAN architecture, a uni-fied control plane with global information about the entire fixed-mobile network status, and the dedicated DL/UL wavelengthchannel for each access node (ONU-eNB), collectively enablethe implementation of a simple and efficient QoS-aware DBAscheme, in which resources are reserved (e.g., queue space andbandwidth) via signaling. Note that the overall process of QoSmapping and support can be further simplified by reducing thenumber of standardized QoS levels for both PON and LTE fromeight to the typical three DiffServ’s classes of services [expe-dited forwarding (EF), assured forwarding (AF), and best effort(BE)], which are commonly and widely used by operators.

VI. MOBILITY MANAGEMENT

Seamless mobility that enables the support of VoIP and otherreal-time IP applications is one of the most important function-


alities of the proposed converged architecture. The convergedarchitecture must support seamless distributed HO proceduresthat conform to the distributed nature of the LTE architecture.In LTE, there is no soft handover support and at each HO theuser context (defines the radio-bearer configurations) and thecoupling between mobility tunnels and radio bearers need to berelocated from one eNB to the other. LTE defines three mobility-states of the UE, LTE-DETACHED, LTE-IDLE, and LTE-AC-TIVE. In LTE-ACTIVE, when a UE roams between two LTEeNB cells, “backward” handover is carried out. Based on mea-surement reports from the UE, the source cell determines thetarget cell and queries the target cell if it has enough resourcesto accommodate the UE [12]–[15]. The target cell prepares radioresources before the source cell commands the UE to handoverto the target cell.

Because data buffering in the DL occurs at the eNB, mech-anisms to avoid data loss during inter-eNB HOs are more crit-ical compared to the 3G architecture where data buffering oc-curs at the centralized RNC and inter-RNC HOs are less fre-quent. The proposed architecture efficiently addresses this issueas described shortly. In this paper, HO is classified into two dif-ferent scenarios, namely, intra-OLT HO and inter-OLT HO. Theformer is a HO between two neighboring eNBs (cells) that arelocated on the same ring and managed by the same OLT (samePON domain), while the latter is between two eNBs located ondifferent adjacent rings, where each eNB is managed by a dif-ferent OLT (each belongs to a different PON domain) but stillmanaged by the same EPC.

A. Registration and Handoff

When a UE enters a domain served by a new PON-RAN, itneeds to register itself to the new domain OLT’s access routerand update the new location in its home subscriber server (HSS).This is done by the new OLT that initiates a location update re-quest to the HSS indicating the change of location to a new OLT.As long as the UE is roaming within the same PON-RAN do-main, it needs not to re-register again. The remaining proceduresfollow the typical LTE registration process.

1) Intra-OLT Handoff: The message sequence diagram ofthe intra-OLT HO procedure between the source and thetarget is shown in Fig. 4. The figure shows both the con-trol plane signaling messages (solid arrows) and the flow of theuser data plane packets (dashed arrows). The UE sends measure-ment reports to the source eNB , which may decide onthe execution of a HO based on these reports. The sourcesends the coupling information and the UE context to the target

requesting the preparation of a HO (HO Request ContextTransfer). The target performs admission control to checkwhether the established QoS bearers of the UE can be accom-modated in the target cell.

Once signals that it is ready to perform the HO (HOAccept), commands the UE to change the radio bearer to

(HO Command). At the same time, to ensure seamlessHO, suspends the RLC/MAC protocols and may start toforward the buffered service data units (SDUs) that have not yetbeen successfully sent to the UE along with all the incoming

Fig. 4. Sequence of the intra-OLT HO procedure between the source ��and the target �� .

SDUs from , if there is any, toward the target . Ac-cording to typical LTE standards, whether SDU forwarding isemployed at all by the eNB is left as a vendor specific imple-mentation detail. However, in the proposed converged architec-ture, it is a simple and straightforward procedure for the source

to forward the SDUs directly to the target as a localLAN traffic on the ring, where the needed direct physical con-nectivity between them exists. However, in LTE, creating a log-ical connectivity between and requires the lengthyprocess of signaling to the MME/S-GW to coordinate the mo-bility-tunnel switch from to .

Next, the UE sends the HO complete message to the target, which is used by the target to verify that it is

the right UE that is accessing the target cell. At that point thetarget can start sending DL data to the UE. For the HOto complete, then signals to inform it that the HOis complete (HO complete) and to update its records with thenew eNB, i.e., to add to the forwarding list for theUE. This means that the scheduler at the OLT will just redirectthe traffic destined to the UE from downstream (connectedto dedicated downstream wavelength serving )to downstream (connected to dedicated downstream wave-length serving ). After receiving the HO com-plete message, first redirects UE’s traffic from toand then removes from the forwarding list of theUE. Then, sends redirect traffic acknowledgement (ACK)to . Upon receiving the ACK, triggers the release ofresources at the source . Finally, signals MME toupdate the UE’s new location.

Clearly, the proposed distributed ring-based unifiedPON-RAN architecture enables the support of a seamlessdistributed intra-OLT HO scheme (inter-eNBs) that has severaladditional significant features compared to the typical LTE’sinter-eNB HO scheme, including: 1) no path switch/setupcommand is needed since the path (mobility tunnels) from EPCto the UE remains unchanged; 2) the EPC is not involved atall except for the simple signaling from to the MME toreport the location update of the UE; 3) re-registration proce-dures to the HSS when the UE moves from to isavoided. It is also avoided as long as the UE roams within thecoverage area served by the cells (eNBs) attached to the ring.


Overall, the proposed architecture significantly reduces thesignaling overhead and HO latency. Furthermore, the proposedHO scheme eliminates the lengthy process of the frequent regis-tration and forwarding path setup, when the UE repeats crossingthe boundary of two adjacent eNBs.

Thus, with very small signaling overheads, the proposed ar-chitecture supports seamless and speedy HO service for the mo-bile nodes when they roam in any PON-RAN domain attached tothe EPC. In addition to directly routing on the ring the bufferedSDUs that have not yet been successfully sent toward the targeteNB from the source eNB during intra-OLT HO, all bidirec-tional upstream data exchange (including VoIP, video, and datasessions) between any two mobile users served by two differenteNBs that are attached to the same ring is also directly routed onthe ring from the source eNB directly to the destination eNB andvice-versa as local LAN traffic, without the direct participationof either the OLT or the EPC. This is significant as the volumeof voice calls and/or multimedia data exchange between localmobile users is substantial. Consequently, a sizable fraction ofthe mobile path switch/setup signaling commands as well as ac-tual local upstream traffic transport and processing are offloadedfrom the EPC to the access nodes (ONUs/eNBs).

2) Inter-OLT Handoff: The first seven steps of the inter-OLTHO are almost identical with those of the intra-OLT HO, asshown in Fig. 4. Starting from step # 8, as shown in Fig. 5, themessage sequence of the inter-OLT HO procedure between thesource located on the first ring and the target lo-cated on the second ring are different. Fig. 5 shows both thesecontrol plane signaling messages (dashed arrows) and mobilitytunnels (solid arrows). First, the UE sends a registration requestto the new OLT once it enters the new domain of thesecond ring. Next, signals MME to coordinate the mo-bility tunnel switch from to and to initiate a lo-cation update request to the HSS, indicating the change of loca-tion to a new OLT. MME then triggers the update at the S-GW toswitch the mobility tunnel, based on the signaling received from

via indicating that radio bearer was successfullytransferred. Once the UE completes the registration to the HSSand new OLT , S-GW will begin to forward packets forit through the new domain access root router at . At thesame time, the HSS notifies the old OLT to cancel thelocation process for this UE. As a result, the old OLT removesUE from its visitor list and releases its resources. Finally,triggers the release of resources at the source .

B. Paging and Efficient Idle Mobility

In idle mode, according to LTE standards, the UE is in power-conservation mode and does not inform the network of eachcell change. In this state, the location of the UE is only knownat the MME and only at the granularity of a few cells, calledthe tracking area (TA). When there is a UE-terminated call, theMME knows the TA in which the UE last registered and pagingis necessary to locate the UE to a cell. This approach, whichregisters to MME/HSS for idle nodes at every few HOs, intro-duces significant signaling overheads and reduces the efficiencyof EPC, especially when the idle node moves quickly. To furtherreduce the registration signaling overhead, the TA is redefined

Fig. 5. Sequence of the inter-OLT HO procedure between the source eNB andthe target eNB.

here to include all the cells (eNBs) attached to the ring (min-imum of 16 cells versus 3–5 cells according to LTE standards).Thus, the idle UE sends a re-registration request to the new OLTwhen it only crosses a PON domain boundary. The new OLTrecords the idle UE in its paging list and reports the locationupdate to the MME/HSS, but it does not allocate resources anddoes not set up a mobility tunnel for the idle UE.

To eliminate the paging signaling overhead, for everyPON-RAN domain, the paging information is broadcasted peri-odically via the downstream Ethernet control frame associatedwith each wavelength channel. When the idle UE moves withinthe same paging domain, it only need to monitor the currentpaging information in the control frame and need not send anymessage to the OLT. If the new OLT receives data destinedfor the idle UE, it buffers the data in its cache and broadcastsa paging request message for the UE within its domain. Uponreceiving the paging message, the UE reports its current lo-cation to the OLT, which then forward the data to the UE.With the application of this paging scheme, the unnecessarysignaling overheads and power waste, which are associatedwith the frequent re-registration for idle mobile nodes, can besignificantly reduced.

Finally, it is important to emphasize that since supportingsimple, efficient, and distributed resilience mechanisms arevital for the viability of the proposed converged access archi-tecture, the distributed control plane can be further utilizedto develop several efficient and cost-effective fully distributedas well as hybrid fault detection and recovery schemes thatprovide the required self-healing mechanisms for the proposedarchitecture. Fault protection techniques for the stand-alonering-based WDM-PON architecture can be found in [38].

VII. CONCLUSION

This paper presents an overview of the most recent researchactivities on NG hybrid FiWi networking architectures withparticular focus on examining the basic differences betweentwo FiWi networking architectural models, namely, the overlayor independent model and the integrated model. A simpleand cost-effective all-packet-based converged fixed-mobileaccess networking solution that enables the true integration ofNG-PON and 4G mobile broadband access technologies intothe envisioned fixed-mobile platform has been presented. Thetrue functional integration of the two access technologies at var-ious networking layers and for several networking operationsand control and management procedures has been outlined


and presented. The integrated model provides the best overallsystem performance in terms of cost-effectiveness, bandwidthutilization, and support for better QoS and efficient seamlessand speedy HO schemes for the mobile nodes when they roamin any PON-RAN domain attached to the EPC. Overall, theintroduction of the OLT as an intermediate node can be used tooffload a considerable portion of the overall system signalingprocessing as well as actual data traffic transport and processingburden form the EPC to the OLT and ONUs, thus alleviatingthe potential AGW’s control-plane scalability bottleneck.

ACKNOWLEDGMENT

The authors would like to thank Dr. A. Shami from the Uni-versity of Western Ontario for numerous useful and insightfuldiscussions as well as technical review of the manuscript.

REFERENCES

[1] IEEE 802.3(a)h EFM Study Group [Online]. Available: http://www.ieee802.org/3/efm/public/index.html, 2001–2004

[2] G. Kramer and G. Pesavento, “Ethernet passive optical network(EPON): Building a next generation optical access network,” IEEEComm. Mag., vol. 40, no. 2, pp. 66–73, Feb. 2002.

[3] L. G. Kazovsky, W._T. Shaw, D. Gutierrez, N. Cheng, and S.-W. Wong,“Next-generation optical access networks,” J. Lightw. Technol., vol. 25,no. 11, pp. 3428–3442, Nov. 2007.

[4] C. Assi, Y. Ye, S. Dixit, and M. A. Ali, “Dynamic bandwidth allo-cation for quality of service over ethernet PONs,” IEEE J. Sel. AreasCommun., vol. 21, no. 9, pp. 1467–1477, Nov. 2003.

[5] S. Sherif, A. Hadjiantonis, G. Ellinas, C. Assi, and M. A. , “A novel dis-tributed ethernet-based PON access architecture for provisioning dif-ferentiated QoS,” J. Lightw. Technol., vol. 22, no. 11, pp. 2483–2497,Nov. 2004.

[6] A. Delowar, R. Dorsinville, M. A. Ali, A. Shami, and C. Assi, “Ring-based local access PON architecture for supporting private networkingcapability,” OSA J. Opt. Netw., vol. 5, no. 1, pp. 26–39, 2006.

[7] J. Kani, F. Bourgart, A. Cui, A. Rafel, M. Campbell, R. Davey, andD. Rodriqgues , “Next-generation PON—Part I: Technology roadmapand general requirements,” IEEE Communication Magazine, vol. 47,no. 11, pp. 43–49, Nov. 2009.

[8] F. J. Effenberger, H. Mukai, S. Park, and T. Pfeiffer, “Next-genera-tion PON—Part II: Candidate systems for next-generation PON,” IEEECommun. Mag., vol. 47, no. 11, pp. 50–57, Nov. 2009.

[9] F. J. Effenberger, H. Mukai, J.-i. Kani, and M. Rasztovitz-Wiech,“Next-generation PON—Part III: System specifications for XG-PON,”IEEE Commun. Mag., vol. 47, no. 11, pp. 58–64, Nov. 2009.

[10] UMTS Evolution, From 3GPP Release 7 to Release 8, HSPA andSAE/LTE UPDATE [Online]. Available: www.3gamericans.org, Jun.2008

[11] I. Ali, A. Casati, K. Chowdhury, K. Nishida, E. Parsons, S. Schmid,and R. Vaidya, “Network-based mobility management in the evolved3GPP core network,” IEEE Commun. Mag., vol. 47, no. 2, pp. 58–66,Feb. 2009.

[12] H. Ekstrom, “QoS control in the 3GPP evolved packet system,” IEEECommun. Mag., vol. 47, no. 2, pp. 76–83, Feb. 2009.

[13] 3GPP TS 23.107, Quality of Service (QoS) Concept and ArchitectureJun. 2005.

[14] 3GPP TS 36.300, Evolved UTRA and Evolved UTRAN, Overall De-scription 2006.

[15] 3GPP TS 23.402, 3GPP Architecture Enhancements for Non-3GPPAccesses May 2007.

[16] N. Ghazisaidi, M. Maier, and C. Assi, “Fiber-wireless (FiWi) accessnetworks: A survey,” IEEE Commun. Mag., vol. 47, no. 2, pp. 160–167,Feb. 2009.

[17] T. Nakasyotani, H. Toda, T. Kuri, and K.-i. Kitayama, “Wavelengthdivision multiplexed millimeter-waveband radio-on-fiber sysem usinga supercontinuum light source,” J. Lightw. Technol., vol. 24, no. 1, pp.404–410, Jan. 2006.

[18] Z. Jia, J. Yu, A. Chowdhury, G. Ellinas, and G. K. Chang, “Simulta-neous generation of independent wired and wireless services using asingle modulator in millimeter-wave-band radio-over-fiber systems,”IEEE Photon. Tech. Lett., vol. 19, no. 20, pp. 1691–1693, Oct. 2007.

[19] J. Zhensheng, J. Yu, G. Ellinas, and G. K. Chang, “Key enabling tech-nologies for optical-wireless networks: Optical millimeter-wave gen-eration, wavelength reuse, and architecture,” J. Lightw. Technol., vol.25, no. 11, pp. 3452–3471, 2007.

[20] C.-T. Lin, J. Chen, P.-C. Peng, C.-F. Peng, W.-R. Peng, B.-S. Chiou,and S. Chi, “Hybrid optical access network integrating fiber-to-the-home and radio-over-fiber systems,” IEEE Photon. Tech. Lett., vol. 19,no. 8, pp. 610–612, Apr. 2007.

[21] J. J. Olmos, T. Kuri, and K. Kitayama, “60-GHz-band 155-Mbps and1.5-Gbps baseband time-slotted full-duplex radio-over-fiber accessnetwork,” IEEE Photon. Tech. Lett., vol. 20, no. 8, pp. 617–619, 2008.

[22] P. S. Henry, “Integrated optical/wireless alternatives for themetropolitan environment,” IEEE Commun. Soc. Webinar, Apr.2007 [Online]. Available: http://www.comsoc.org/webinar

[23] S. Sarkar, S. Dixit, and B. Mukherjee, “Hybrid wireless-optical broad-band-access network (WOBAN): A review of relevant challenges,” J.Lightw. Technol., vol. 25, no. 11, pp. 3329–3340, Nov. 2007.

[24] W. Shaw, S.-W. Wong, N. Cheng, K. Balasasubramanian, X. Zhu, M.Maier, and L. Kazovsky, “Hybrid architecture and integrated routingin a scalable optical-wireless access network,” J. Lightw. Technol., vol.25, no. 11, pp. 3443–3451, Nov. 2007.

[25] S. Sarkar, B. Mukherjee, and S. Dixit, “Optimum placement of mul-tiple optical network units (ONUs) in optical-wireless hybrid accessnetworks,” in IEEE/OSA Proc. Opt. Fiber Commun. Conf., Anaheim,CA, Mar. 2006, p. 3.

[26] S. Sarkar, H.-H. Chen, D. Dixit, and B. Mukherjee, “DARA: Delayaware routing algorithm in a hybrid wireless-optical broadband ac-cess network (WOBAN),” in IEEE Proc. Int. Conf. Commun. (ICC),Glasgow, U.K., Jun. 2007, pp. 2480–2484.

[27] S. Sarkar, H.-H. Yen, S. Dixit, and B. Mukherjee, “RADAR: Risk-and-delay aware routing algorithm in a hybrid wireless-optical broadbandaccess network (WOBAN),” in IEEE/OSA Proc. Opt. Fiber Commun.Conf., Anaheim, CA, Mar. 2007, pp. 1–3.

[28] J. J. Olmos, T. Kuri, T. Sono, K. Tamura, H. Toda, and K.-i. Kitayama,“Wireless and optical-integrated access network with peer-to-peer con-nection capability,” IEEE Photon. Technol. Lett., vol. 20, no. 13, pp.1127–1129, Jul. 2008.

[29] Y. Luo, N. Ansari, T. Wang, M. Cvijetic, and S. Nakamura, “A QoS ar-chitecture of integrating GEPON and WiMAX in the access network,”in Proc. IEEE Sarnoff Symp., Princeton, NJ, Mar. 2006, pp. 1–4.

[30] Y. Luo, T. Wang, S. B. Weinstein, M. Cvijetic, and S. Nakamura, “In-tegrating optical and wireless services in the access network,” in Proc.IEEE/OSA National Fiber Optic Engineers Conference (NFOEC),Anaheim, CA, Mar. 2006.

[31] M. Bakaul, N. Nadarajah, A. Nirmalathas, and E. Wong, “Internet-working VCSEL-based hybrid base station towards simultaneous wire-less and wired transport for converged access network,” IEEE Photon.Tech. Lett., vol. 20, no. 8, pp. 569–571, Apr. 2008.

[32] N. Nadarajah, M. Bakaul, and A. Nirmalathas, “Laser-free inter-net-working hybrid base stations towards convergence of wireless andwired access networks,” Electron. Lett., vol. 43, no. 8, pp. 472–474,Apr. 2007.

[33] E. Wong, C. Lim, A. G. Prasanna, K. L. Lee, A. Nirmalathas, M.Ortsiefer, and C. Neumeyr, “Simplified base station configuration forfiber-wireless applications,” presented at the presented at the 33 Eur.Conf. Optical Communication (ECOC), Berlin, Germany, Sep. 2007.

[34] K. Yang et al., “Convergence of ethernet PON and IEEE 802.16 broad-band access networks and its QoS-aware dynamic bandwidth allocationscheme,” IEEE J. Sel. Areas Commun., vol. 27, no. 2, pp. 101–116, Feb.2009.

[35] G. Shen and R. S. Tucker, “Fixed mobile convergence architectures forbroadband access: Integration of EPON and WiMAX,” IEEE Comm.Mag., vol. 45, no. 8, pp. 44–50, Aug. 2007.

[36] H. Erkan, ASM D. Hossain, R. Dorsinville, M. A. Ali, A. Hadjiantonis,G. Ellinas, and A. Khalil, “A novel ring-based WDM-PON access ar-chitecture for efficient utilization of network resources,” in IEEE Proc.Int. Conf. Commun. (ICC), Beijing, China, May 2008, pp. 5175–5181.

[37] 4G LTE, 2007 [Online]. Available: http://www.mobileeurope.co.uk/features/112515/4G_LTE.html

[38] H. Erkan, G. Ellinas, A. Hadjiantonis, R. Dorsinville, and M. A. Ali,“Native ethernet-based self-healing WDM-PON local access ringarchitecture: A new direction for supporting simple and efficientresilience capabilities,” presented at the presented at the IEEE Int.Conf. Communications (ICC), Cape Town, South Africa, May 2010.


Mohamed A. Ali received the M.S. and Ph.D. de-grees in electrical engineering from The City Collegeof the City University of New York, in 1985 and 1989,respectively.

In 1989, he joined the Faculty of Electrical Engi-neering at The City College of the City Universityof New York, New York, where he is currently aProfessor. Since 1995, he has been an IT Consultantfor several major telecom carriers in the USA. Heis the author or coauthor of more than 120 refereedjournal papers, invited talks, conference proceed-

ings, and several book chapters. His research interest include the general areaof telecommunications networking architecture and technology encompassingbroadband wired/wireless access, IP/multiprotocol label switching-basedlayer 1/2/3 enterprise metropolitan area network and virtual private networks,fourth-generation (4G) mobile WiMAX and cellular long-term evolutionsystems, passive optical networks, long-haul wavelength-division-multiplexingand time-division-multiplexing/synchronous optical network-based opticaltransport networks, etc.

Prof. Ali received the National Science Foundation Faculty Career Develop-ment Award.

Georgios Ellinas received the B.S., M.S., M.Phil.,and Ph.D. degrees in electrical engineering from Co-lumbia University, New York.

He is currently an Associate Professor at the De-partment of Electrical and Computer Engineering,University of Cyprus, Nicosia. From 2002 to 2005,he was an Associate Professor of electrical engi-neering at the City College of the City Universityof New York. During 2000–2002, he was a SeniorNetwork Architect at Tellium, and from 1998 to2000, he was a Senior Research Scientist in Telcordia

Technologies’ (formerly Bellcore) Optical Networking Research Group. Heis the author or coauthor of two books on optical networking and more than120 journal and conference papers and book chapters. He holds 29 patents onoptical networking. His research interests include optical architectures, routingand wavelength assignment algorithms, fault protection/restoration techniquesin mesh optical networks, optical access networks, hybrid optical-wirelessaccess networks, and complex networks.

Hasan Erkan received the Ph.D. degree in electrical engineering from theGraduate School and University Center of the City University of New York,New York, in 2008.

He is currently with the Department of Engineering, Maritime College, StateUniversity of New York. His research interests include optical network archi-tectures and optical access networks with a focus on wavelength-division mul-tiplexing-passive optical networks.

Antonis Hadjiantonis received the Ph.D. degreein electrical engineering from the Graduate Schooland University Center of the City University of NewYork, New York, in 2006.

He is currently an Assistant Professor at theDepartment of Engineering, University of Nicosia,Nicosia, Cyprus. He is also a Senior Researcher atthe Cyprus Academic Research Institute, where heis involved in various funded research projects. Hisresearch interests include the vertical integration inmultilayer networking environments, routing and

signaling algorithms, the first/last mile, and fixed-mobile convergence.Dr. Hadjiantonis received the prestigious Carell Dissertation Fellowship

Award for his outstanding research in optical networking.

Roger Dorsinville received the Ph.D. degree inphysics from Lomonosov State University, Russia.He is currently Professor of Electrical Engineeringand Chair of the EE Department at The City Collegeof the City University of New York (CUNY). Hisresearch interests include nonlinear optical proper-ties of composite materials, fiber lasers, all opticalcomponents for communication systems and passiveoptical network architectures.

On the Vision of Complete Fixed-Mobile Convergence

Documents