1 Hybrid Wireless-Optical Broadband Access Network (WOBAN): Prototype Development and Research Challenges Pulak Chowdhury, Suman Sarkar, Glen Kramer, Sudhir Dixit, and Biswanath Mukherjee Abstract—Hybrid wireless-optical broadband access net- (“anytime-anywhere”) broadband access to satisfy fu-work (WOBAN) is emerging as a promising technology ture customer demands. Therefore, a novel cross-domainto provide economical and scalable broadband Internet network paradigm – Wireless-Optical Broadband Accessaccess. In this cross-domain network architecture, end- Network (WOBAN) – which is an optimal combinationusers receive broadband services through a wireless mesh of high-capacity optical backhaul and untethered wirelessfront-end which is connected to the optical backhaul viagateway nodes. In this article, we present the architecture access, is proposed in the literature .and functional characteristics of a WOBAN prototype built WOBAN shows excellent promise for future accessin the Networks Lab. at UC Davis. We cite some research networks. This cross-domain network architecture con-challenges on hybrid networks based on our experimental sists of an optical backhaul (e.g., a Passive Optical Net-observations. work (PON)) and wireless access in the front-end (e.g., Index Terms—WOBAN, Prototype, Hybrid, Cross- WiFi and/or WiMAX). In WOBAN, a PON segment startsDomain, Optical-Wireless. from the telecom Central Ofﬁce (CO) with an Optical Line Terminal (OLT) at its head end. Each OLT can drive several Optical Network Units (ONU), and each ONU I NTRODUCTION can support several wireless routers of the wireless front- During the past decade, the backbone network has end in WOBAN. The wireless routers directly connectedexperienced enormous growth in capacity and reliability, to the ONUs are called as wireless gateways. The wire-mainly due to major development efforts in the area of less front-end also consists of other wireless routers tooptical networking. During the same time, bandwidth provide end-user connectivity. Therefore, the front-enddemands of technology-savvy end users for broadband of a WOBAN is effectively a multi-hop Wireless Meshservices such as “quad-play” (voice, video, Internet, and Network (WMN) which is connected to the high-capacitywireless) and media-rich applications have also increased PON segment in the back-end, creating a cross-domainat an unprecedented rate. However, the access network integrated network architecture.(commonly referred to as the “last-mile” network) still There is another related architecture, known as Radio-remains a bottleneck for providing bandwidth-intensive Over-Fiber (ROF), where radio signals can be effectivelyservices to customers. Legacy access technologies (such carried over an existing optical ﬁber infrastructure usingas Digital Subscriber Line (DSL) and Cable Modem “Hybrid Fiber Radio” (HFR) technology . ROF deals(CM)) will not be able to carry the high volume of with the communication challenges of sending radio sig-trafﬁc generated by emerging applications such as video- nals over ﬁber whereas WOBAN focuses on the network-on-demand (VoD), interactive gaming, or duplex video- ing aspects of the wireless-optical converged architecture.conferencing. Thus, future access technologies should In this article, we present the experiences gatheredprovide high capacity and operational efﬁciencies along during a WOBAN prototype development, and discusswith mobility support and untethered access to users in a future research issues to improve the performance andcost-effective manner. design of this hybrid network. We provide detailed proto- Optical-ﬁber-based technologies (e.g., ﬁber-to-the- type development procedures and introduce some of thehome (FTTH), ﬁber-to-the-building (FTTB), ﬁber-to-the- challenges involved in the development. The WOBANcurb (FTTC)) are well suited to support integrated high- prototype serves as the experimental setup for variousbandwidth digital services, and can alleviate bandwidth access network protocols and data dissemination tech-bottlenecks. The next generation of access networks is niques; and it features programmability, resource sharing,therefore promising to deploy optical ﬁber all the way to and slice-based experimentation. We believe that thisthe customer premises. However, laying ﬁber infrastruc- prototyping effort will lead us to identify and addressture to all end-users incurs signiﬁcant cost. Furthermore, several practical concerns that WOBAN may encounterusers also desire untethered access, especially if they are in future.mobile. Wireless technologies can support mobility and The remainder of this article is organized as follows.untethered access. Unfortunately, wireless access is con- We ﬁrst present related prototyping efforts on hybridstrained due to limited bandwidth. Therefore, combining cross-domain networks in the literature. We then presentthe complementary features of these two technologies the WOBAN prototype architecture, its distinguishing(optical and wireless) can potentially provide ubiquitous features, and its development procedure. Experimental
2 TABLE Iresults are demonstrated and discussed in the following WOBAN P ROTOTYPE C OMPONENTS AND T HEIR S PECIFICATIONS .section. Then, we elaborate on future research challengesof WOBAN. Finally, concluding remarks are provided. Components Interface/Port OLT • Client Side: One EPON port • Network Side: One 100/1000 R ELATED D EVELOPMENT E FFORTS Base-T Ethernet port (for RoI (Rest-of-the-Internet)) This section brieﬂy reviews other testbeds/prototypes ONU • Client Side: Two 10/100 Base-Tdeveloped for hybrid wireless-optical networks research. Ethernet ports (to drive 802.11g Hu et al.  have developed a testbed for an Optical- routers)Wireless Integration (OWI) infrastructure. They imple- • Network Side: One EPON portmented SONET/WDM, popular in core optical networks, (to connect OLT)for the optical part and WiMAX (IEEE 802.16) for Optical Splitter 1:8 power splitterbroadband wireless access. The edge node between two 802.11g Router • Client Side: One radio portnetworks interfaces the WiMAX base station and SONET • Network Side: 10/100 Base-Twith a direct conversion between the protocol stacks of Ethernet portthe optical and wireless segments. Grid Reconﬁgurable Optical and Wireless Network Clients Laptops, PDAs, etc.(GROW-Net)  is another hybrid wireless-optical net-work which consists of an “Infrastructure” based WMNin the front-end and a reconﬁgurable, high-capacity, point- Architectureto-multipoint PON optical backhaul. To demonstrate theperformance of the proposed optical backbone reconﬁg- Figure 1 shows the architecture of WOBAN prototypeuration scheme in GROW-Net, the authors of  de- developed in the Networks Research Laboratory at UCveloped only an optical experimental testbed based on Davis.commercially-available devices. This testbed is dedicatedto optical backhaul reconﬁguration experiments. The wireless routers form the WOBAN front-end and Jia et al.  have developed a testbed for Radio-Over- connect to the end users (who can be scattered over theFiber (ROF) experiments. The testbed has two segments – geographic area served by the WOBAN and who are not(a) Central Station (CS) and (b) Base Station (BS) – and shown in Fig. 1). These wireless routers (IEEE 802.11g)it consists of optical transmission equipments. The main support data rates up to 54 Mbps. Several designatedpurpose of this testbed is to illustrate how wireless signals routers are conﬁgured to have Gateway capabilities (bycan be carried over ﬁber. This testbed demonstrates the loading appropriate open source ﬁrmware) and each suchfeasibility of a full-duplex ROF system based on optical Gateway is connected to an ONU via a 10/100 Base-Tcarrier suppression and reuse for future optical/wireless Ethernet port. The wireless routers are placed with annetworks. effective distance of 50-60 meter between pairs. Two OLTs (Optical Line Terminal) emulate the func- tionality of the telecom Central Ofﬁce (CO) of the general I MPLEMENTING WOBAN P ROTOTYPE WOBAN architecture. Each OLT can drive several ONUs using an optical splitter. The OLTs and ONUs are con- In this section, we discuss the logistics (resources nected through Ethernet PON (EPON) ports. The OLTsneeded for prototype development), WOBAN architec- are connected to the Rest of the Internet (ROI) using theture, features, and detailed development procedure. campus-wide backbone network at UC Davis. The prototype architecture is divided into three planes:Resources Needed (a) Control Plane, (b) Data Plane, and (c) Management Plane. The Control Plane is used to deﬁne different Table I summarizes various device speciﬁcations used control features of the nodes in the WOBAN prototype.in our prototype. All these devices are commercially The Data Plane conﬁgures routing and different data-available off-the-self devices and can be used effectively transfer scenarios, and collects measurement data forto build a fully-functional and reasonable-sized prototype. different experiments. The Management Plane is used We use open source ﬁrmwire OpenWRT1 to develop for remote access and programmability of the prototypethe reconﬁgurable wireless routers and gateways. nodes. The WOBAN Network Operations Center (NOC) (see Fig. 1) is responsible for the management of all these 1 “OpenWrt”, http://www.openwrt.org/, 2009. planes.
3 ne P la e nt g em lane P lan e a aP M an D at tr o l C on Management v Portal v Controller v Wireless Link Linux Box 54 Mbps WOBAN NOC 10/100 Base T RJ45 Router Gateway 1 ONU 1 100M/1G Wireless Routers and Base T Fiber Gateways are configured differently OLT 1 1:8 RJ45 for their respective Optical Linux Box Rest of the Internet functionalities Splitter ONU 2 (UCDavis Network) 10/100 100M/1G Base T Base T Fiber Gateway 2 Router RJ45 OLT 2 1:8 Optical ONU 3 Splitter WOBAN CO 10/100 Wireless Link Base T 54 Mbps Gateway 3 Linux Box Control, Data, and Management Planes at WOBAN Each OLT drives different Each ONU and its attached Linux Box Wireless Routers have distributed control: link formation NOC parts of WOBAN emulate the operation of an ONU and routingFig. 1. WOBAN prototype architecture.Distinguishing Features box and the ONU programmability can be emulated by gluing a separate “Linux box” with each ONU. The WOBAN prototype has several distinguishing fea- • The prototype is reconﬁgurable and provides self-tures which are different from other related prototypes organizing and self-healing properties. The reconﬁg-(, , ) reported in the literature, as follows. urability is performed by Layer-2 (L2) connectivity • To the best of our knowledge, this is the most inte- and intelligent routing. grated wireless-optical hybrid network testbed. Other • Power consumption of the wireless nodes is very low testbeds have only a small number of nodes and (1-2.5 watts/router). As the wireless mesh constitutes have been used as proof of concepts. On the other a large part of the prototype, the overall power hand, WOBAN prototype features programmability, consumption is also low. self organization, and slice-based experimentation. • The WOBAN prototype is large enough to demon- strate its useful properties, e.g., two OLTs can Development Procedure demonstrate fault-tolerance properties of WOBAN so that, if one OLT breaks, the other parts of the Here, we present deployment issues related to different WOBAN can “self organize” themselves to still planes in the WOBAN prototype and show how they are carry the affected trafﬁc through the other opera- addressed during the deployment phase. tional parts of the WOBAN. The self-organization Control-Plane Issues: property of WOBAN also holds for (1) other failure • Topology Creation/Connectivity: The optical seg- types, e.g., ONU failure, ﬁber cut, wireless router ment of the WOBAN prototype has a static topology failure, etc. and (2) optimal routing. initially as connections between nodes are wired. • The deployment and management cost of WOBAN The wireless segment uses proactive routing (namely prototype is low as it is built from highly-customized Optimal Link State Routing (OLSR) in our proto- off-the-shelf components, open sources, and indige- type) to create a “self organizing” topology where, nous software. in case of a router failure, nodes can redirect trafﬁc • The front-end can be set up as a plug-and-play to the nearby active routers. If a failure occurs in the wireless mesh. optical segment, dynamic protection scheme can be • The prototype nodes feature programmability. The applied for “self-healing”. open source ﬁrmware provides the programmability • Dynamic Bandwidth Allocation (DBA): The op- in the wireless routers. The programmability of OLT tical part of the WOBAN prototype uses Ethernet can be performed by using the craft port in the OLT PON (EPON) as the basic technology. In EPON,
4 the Ethernet functionality is emulated by a Layer- We can also implement the virtual slicing feature 2 signalling mechanism, called Multi-Point Control where the physical resources of WOBAN nodes Protocol (MPCP)  that would allow the OLT to can be shared among experiments. Time-Division assign the bandwidth dynamically among ONUs. We Multiplexing (TDM) based virtual slicing is very can use hierarchical MPCP-based protocol in two challenging to implement . Further research is levels (OLT-to-ONUs and ONU-to-Gateways) cou- required to deploy such features in the prototype. pled with Layer-2 signaling (Gateways-to-Routers) for DBA, and thereby achieve stronger wireless- optical integration. Overview of this kind of protocol E XPERIMENTAL I LLUSTRATIONS is given in a later section.• Programmability: An important aspect of the Here, we present experimental results collected from WOBAN prototype nodes is their programmability. the WOBAN prototype for various applications (Data, Experimental testbed researchers should be able to Voice-over-IP (VoIP), and Video-on-Demand (VoD)). create, modify, and test their protocols on the proto- type. In our WOBAN prototype, we create a simple remote-access-based programmability platform for Experimental Setup the wireless nodes (gateways/routers). This platform provides programmability at each layer of the IEEE Figure 2 shows the setup for different experiments on 802.11 protocol stack. The OLT DBA mechanism WOBAN prototype. The wireless front-end of WOBAN (Layer-2 signalling) can also be programmed using should have distributed control to exhibit self-healing the craft port installed in the OLT box. For ONU, we and self-organization properties. Therefore, we use IEEE can emulate the programmability by gluing a “Linux 802.11 basic Distributed Coordination Function (DCF) box” with each of them. for medium sharing. IEEE 802.11 Point CoordinationData-Plane Issues: Function (PCF) is only suitable in wireless “infrastruc- ture” mode, hence is not feasible for WOBAN wireless• Routing: Proactive routing such as Optimal Link mesh front-end. IEEE 802.11e-based enhanced coordina- State Routing (OLSR) is used in the wireless mesh tion functions for better QoS performance have not been and Layer-2 static routing is used in the optical considered in our experiments as this standard is relatively part of the WOBAN prototype. Dynamic routing new and is still in the development phase. Transmission protocols such as OLSR waste signiﬁcant amount of rate of wireless routers is set to 54 Mbps. wireless bandwidth for periodic link-state updates. From our prototype experience, we ﬁnd that static In all the experiments, background trafﬁc load is gen- routing can perform better compared to a dynamic erated using software-based trafﬁc generators. We run approach in a WOBAN-type network architecture. our experiments under no background load to heavy One such proposal is discussed below. background load to ﬁnd out the effects of background• Conﬁgurations: Prototype nodes can be conﬁgured trafﬁc on different applications. In all the experimental for different experiments. These data-transfer conﬁg- setups, one end (server/client) of a connection is located urations facilitate us to obtain experimental data for in the RoI, and the other end (wireless client) is connected various applications on the WOBAN prototype. to the wireless mesh through multiple hops. Background• Measurement: Network protocol analyzers (e.g., trafﬁc also ﬂows between these two ends so that all tcpdump, Wireshark2 , etc.) are used to collect and the links of a experimental connection experience some analyze network statistics from various experiments. external trafﬁc load.Management-Plane Issues: The quality of the wireless channels varies randomly• Remote Access: In the WOBAN prototype, we use across the experiments due to different interference fac- remote access interfaces to download our own code tors in our environment. This inherent randomness of inside the nodes and run the experiments. Wireless wireless channels may have impacts on accumulated nodes are connected with the NOC through wireless results. The impact of wireless channel quality on the interfaces, and optical nodes are connected through performance is not studied in these experiments. We craft ports. mainly focus on various applications’ performance under• Network Slicing: To share the WOBAN testbed re- random wireless environments. Our results indicate that, sources among several experiments, currently phys- as the number of wireless hops increases, various perfor- ical slicing is used. In physical slicing, resources mance quality measures decrease, due to bottleneck in the are physically divided among different experiments. wireless mesh. Therefore, our accumulated results present the performance of different applications by varying the2 “Wireshark”, http://www.wireshark.org/, 2009. number of wireless hops.
5 Video over 3 wireless hops Wireless Link Video Client DatSerer Veo erer Gateway 1 a v ud s d s u er no.s c s du d Sv c a.c.ucav.ed s ra c.udav.e i i ONU 1 i Data over 4 Rest of the Internet OLT 1 Splitter wireless hops (UCDavis Network) ONU 2 VoIP over 2 wireless hops Data Client Gateway 2 OLT 2 Splitter VoIP Client ONU 3 VoIP Client Wireless Link Gateway 3 Optical Wireless Backhaul Front-EndFig. 2. WOBAN prototype experimental setup. Throughput vs. Wireless Hop CountResults 3 Without Background Traffic Background Traffic 1.5 Mbps Background Traffic 3.0 Mbps Data: We start with data-transfer applications such as 2.5secure ﬁle transfer (viz., sftp or winscp). In our experi- Throughput (Mbps) 2ments, transferred ﬁle size is 76 MBytes. Figure 3 showsthe data-transfer application’s end-to-end throughput. As 1.5expected, with increasing number of wireless hops, end-to-end throughput decreases signiﬁcantly. Furthermore, 1presence of background trafﬁc decreases the throughput. 0.5 Voice-over-IP (VoIP): Next, we present the VoIP end-to-end performance. We use skype as the VoIP applica- 0 1 2 3 4 Wireless Hop Counttion. Figure 4 presents different performance measuresfor skype-based experiments. As the number of wireless Fig. 3. Data-transfer throughput.hops increases, both packet-loss rate and jitter increase,resulting in degraded voice quality. Voice quality alsodegrades with the increase of background trafﬁc load. Streaming Server3 as VoD server and VLC Player4 asWe use the performance metric of Mean Opinion Score client for our video experiments. In this real-time video(MOS)  to measure the subjective voice quality. MOS streaming scenario, the VoD server broadcasts the videogives a numerical indication of the perceived voice quality and the client plays the broadcasted streaming video. Theat the receiver end. MOS is expressed in one number, broadcasted streaming video ﬁle is 30 sec. in duration,from 1 to 5, 1 being the worst and 5 being the best. A 640×480 pixels in size, and encoded at 500 kbps. Figuresgroup of regular VoIP users were asked to give a score 5(a) and 5(b) show the corresponding packet-loss ratebetween 1 to 5 after experiencing the voice quality in and jitter, respectively, with number of wireless hops.different experiments. Then, the mean is calculated to Figures 5(c)-5(f) (screen shots taken at 17 sec. of thedetermine the MOS for different experimental setups. By video streaming on the client side) show the qualitativecomparing the VoIP performance measures, it is evident video streaming performance with different number ofthat packet-loss rate increases (hence voice quality (or wireless hops. In these ﬁgures, the background trafﬁc isMOS) decreases) with the number of wireless hops. As assumed to be moderate (1.5 Mbps). As the number ofexpected, too many wireless hops will not help to improve wireless hops increases and as expected, the video packet-the WOBAN performance. 3 “Darwin Streaming Server”, http://developer.apple.com/opensource- Video-on-Demand (VoD): Performance measures for /server/streaming/index.html, 2009.video transmission are presented in Fig. 5. We use Darwin 4 “VLC Player”, http://www.videolan.org/vlc/, 2009.
6 Packet Loss Rate vs. Wireless Hop Count MOS vs. Wireless Hop Count Jitter vs. Wireless Hop Count 5 30 220 Without Background Traffic Without Background Traffic Without Background Traffic Background Traffic 1.5 Mbps Background Traffic 1.5 Mbps 200 Background Traffic 1.5 Mbps Background Traffic 3.0 Mbps 25 Background Traffic 3.0 Mbps Background Traffic 3.0 Mbps 180 4 Packet Loss Rate (%) 20 160 Jitter (ms) MOS 140 15 3 120 10 100 2 80 5 60 0 40 1 1 2 3 4 1 2 3 4 1 2 3 4 Wireless Hop Count Wireless Hop Count Wireless Hop Count (a) (b) (c)Fig. 4. VoIP performance: (a) Packet-loss rate, (b) Jitter, and (c) Mean Opinion Score (MOS). Packet−Loss Rate vs. Wireless Hop Count Jitter vs. Wireless Hop Count 40 300 Without Background Traffic Without Background Traffic Background Traffic 1.5 Mbps Background Traffic 1.5 Mbps 35 Background Traffic 3.0 Mbps Background Traffic 3.0 Mbps 250 30 Packet−Loss Rate (%) 25 200 Jitter (ms) 20 15 150 10 100 5 0 50 1 2 3 4 1 2 3 4 Wireless Hop Count Wireless Hop Count (a) (b) (c) (d) (e) (f)Fig. 5. Video streaming performance: (a) Packet-loss rate, (b) Jitter, (c) Original video, and at 1.5 Mbps background trafﬁc video quality (d)After one wireless hop, (e) After two wireless hops, and (f) After three wireless hops.loss rate increases, and the video quality deteriorates. • Many wireless hops do not help. But intelligentTill two wireless hops, we can receive decent quality Gateway placement in the wireless mesh may helpof video. After three hops, the video is blurred (Fig. to reduce the number of wireless hops, and improve5(f)), and after four hops only a blank screen shows up the overall WOBAN performance. We can also putin the video client. A heavily-congested network also more Gateways in the mesh to decrease the numbersigniﬁcantly affects the quality of video transmission. of wireless hops.Therefore, the wireless mesh front-end of the WOBAN • Intelligent channel assignment in the wireless meshshould not have many wireless hops if it has to provide can help to improve performance. We found that,quality broadband services to end users. during our mesh setup, if channel 1 of the 2.4-GHz band is assigned to the wireless routers, we can getCritical Observations better results compared to assigning channel 6. This We accumulate the following observations from our is due to several other interfering routers (outsideWOBAN prototyping procedure and experiments. of our WOBAN) near the mesh setup working on
7 channel 6. All the results presented in this article The optical segment of WOBAN already uses MPCP- have been collected using channel 1. based DBA, namely Interleaved Polling with Adaptive • A dynamic link-state routing protocol such as OLSR Cycle Time (IPACT) . Therefore, one can develop a wastes a lot of wireless bandwidth. As the WOBAN hierarchical MPCP-based L2 routing for WOBAN (multi- front-end is a relatively static mesh and a small point control for an OLT to its downstream ONUs and number of wireless hops is needed for improved for an ONU to its downstream Gateways). The idea of performance, the WOBAN mesh performance can L2 routing can be extended in the optical segment (till be improved by using static routing. the Gateways) so that it ﬁts with the wireless mesh • Wireless nodes near a Gateway carry more trafﬁc architecture with one ONU driving multiple Gateways compared to distant ones. Therefore, the memory (similar to the case where one OLT drives multiple and processing power of these “closer” nodes should ONUs). The wireless mesh will use a spanning tree for be higher. Moreover, from prototyping viewpoint, L2 routing. This approach is consistent with the idea of current processing power and memory of off-the- end-to-end L2 capability of WOBAN. shelf wireless routers will not be sufﬁcient for virtual slicing (where several experiments are running on the TDM MAC for Wireless same physical resources). Traditional wireless mesh uses collision-based MAC • As the optical segment of the WOBAN prototype protocols. Our current deployment based on IEEE uses a TDM-based Medium Access Control (MAC) 802.11g wireless routers uses Carrier Sense Multiple scheme, for better wireless and optical integration Access with Collision Avoidance (CSMA/CA) MAC pro- and for improved performance, a TDM-based MAC tocol. From our testbed experience, it is evident that would be a better choice for the wireless mesh. CSMA/CA poses a hindrance on the limited wireless • For video transmission, the standard MAC protocol capacity. From the literature, we ﬁnd that a TDM-based is not sufﬁcient. The MAC layer should be able to MAC protocol can improve the capacity of the wireless distinguish and prioritize between video frames and mesh. Furthermore, as we envisioned for a L2 routing other trafﬁc for better video performance. approach earlier, a TDM-based MAC will also be con- • Although a wireless node can have a theoretical max- sistent with a L2 routing protocol. Therefore, a TDM- imum capacity of 54 Mbps, due to interference and based MAC protocol for the wireless mesh will lead to other surrounding interference, the wireless capacity the seamless integration of both optical and wireless seg- achieved is very low. ments of WOBAN. Other MAC protocols like Orthogonal • Routing in the wireless mesh without considering the Frequency Division Multiplexing (OFDM) combined with optical segment’s trafﬁc condition does not help, and TDM can also be considered in the future to improve vice versa. Therefore, an integrated routing approach wireless capacity. will help to improve WOBAN performance. Improve Flexibility in WOBAN Architecture R ESEARCH C HALLENGES Existing PON technologies do not exhibit sufﬁcient fault tolerance and self-organization capabilities. In case In this section, we discuss some research challenges of OLT, ONU, or wireless gateway failures in a WOBAN,which we have accumulated from the experience gathered we need to redirect the trafﬁc to other live nodes. Thefrom our WOBAN prototype development. self-organization and fault-tolerant properties of WOBAN should ensure this ﬂexibility. Moreover, when an ONULayer-2 Integrated Routing gets congested due to heavy load, we need to perform Current deployment of WOBAN assumes separate load shifting and load balancing so that the network’sdata-transfer techniques for optical and wireless seg- health is ensured.ments. In the optical part, we use MPCP-based DynamicBandwidth Allocation (DBA), whereas the wireless mesh Hierarchical Architectureuses Layer-3 routing, namely OLSR. So, current WOBAN From our experimental observations, it is clear thatdeployment employs a loosely-integrated network ar- wireless Gateways and routers near a Gateway carry morechitecture and control. Layer-3 routing in the wireless trafﬁc compared to routers which are far away from amesh also poses signiﬁcant overhead on the network. To Gateway. Therefore, the routers in the vicinity of theprovide seamless integration of the optical and wireless Gateway and the Gateway itself should be well-equippedsegments, and to reduce Layer-3 processing overheads, with high-capacity wireless resources. The capacity ofan interesting alternative is an integrated Layer-2 (L2) wireless routers can be increased using technologies suchrouting protocol which can efﬁciently route trafﬁc through as multiple radios, directional antenna, and Multiple Inputall segments of WOBAN. Multiple Output (MIMO), etc.
8 C ONCLUSION In this article, we showed how to build a prototypefor a novel, high-bandwidth future access network tech-nology, named WOBAN. This technology is envisionedto satisfy future bandwidth demand of technology-savvycustomers in a cost-effective manner, and it can be anattractive solution for future “last-mile” access networks.We demonstrated the performance of several typical ap-plications such as data transfer, voice, and video over ourWOBAN prototype. We observed that too many wirelesshops degrade the application performance, particularlyfor video. Future research challenges accumulated fromour prototyping experiences were also illustrated. TheWOBAN prototype will be instrumental to develop, test,and analyze the performance of hybrid network protocols.This programmable and conﬁgurable access architecturewill facilitate future experimental, hybrid, and cross-domain networking research. R EFERENCES S. Sarkar, S. Dixit, and B. Mukherjee, “Hybrid Wireless-Optical Broadband Access Network (WOBAN): A review of relevant challenges,” IEEE/OSA Journal of Lightwave Technology, Spe- cial Issue on Convergence of Optical Wireless Access Networks, vol. 25, no. 11, Nov. 2007, pp. 3329–3340. IEEE/OSA Journal of Lightwave Technology, Special Issue on Convergence of Optical Wireless Access Networks, Nov. 2007. J. Hu, D. Qian, H. Yang, T. Wang, S. Weinstein, M. Cvijetic, and S. Nakamura, “Triple play services over a converged opti- cal/wireless network,” Proc. OFC ’06, Anaheim, CA, Mar. 2006. W.-T. Shaw, S.-W. Wong, N. Cheng, K. Balasubramaian, X. Qing, M. Maier, and L. G. Kazovsky, “Hybrid architecture and integrated routing in a scalable optical-wireless access network,” IEEE/OSA Journal of Lightwave Technology, vol. 25, no. 11, Nov. 2007, pp. 3443–3451. Z. Jia, J. Yu, and G.-K. Chang, “A full-duplex radio-over-ﬁber system based on optical carrier suppression and reuse,” IEEE Pho- tonics Technology Letters, vol. 18, no. 16, Aug. 2006, pp. 1726– 1728. G. Kramer, Ethernet Passive Optical Networks, McGraw-Hill, USA, 2005. G. Smith, A. Chaturvedi, A. Mishra, and S. Banerjee, “Wireless virtualization on commodity 802.11 hardware,” Proc. WINTECH 2007, Montreal, Quebec, Canada, Sept. 2007, pp. 75–81. “ITU-T P.800.1 Mean Opinion Score (MOS).” http://www.itu.int/itudoc/itu- t/aap/sg12aap/history/p800.1/p800.1.html, 2009. G. Kramer, B. Mukherjee, and G. Pesavento, “Interleaved Polling with Adaptive Cycle Time (IPACT): A dynamic bandwidth dis- tribution scheme in an optical access network,” Photonic Network Communications, vol. 4, no. 1, Jan. 2002, pp. 89–107.