INVITED PAPER Evolution of Optical Access Networks: Architectures and Capacity Upgrades This overview of passive optical network technologies covers their role from an architectural perspective; methods to upgrade capacity are outlined and evolution strategies are discussed for efficient migration. By Glen Kramer, Marilet De Andrade, Rajesh Roy, and Pulak Chowdhury ABSTRACT | Passive optical network (PON) is one of the most successful broadband access architectures being deployed worldwide. PONs provide high capacity, increased reach, and low-power consumption at a very reasonable cost, on par with the cost of DSL deployments today. This paper provides an overview of present and emerging PON technologies, and discusses PON’s important role in the evolution of optical access from the architectural perspective. While describing the evolution of optical access architecture, we present two important integration options: optical+wireless access integration and metro+access integration. Potential PON capacity upgrades are discussed with special emphasis on achieving a seamless upgrade. We evaluate different PON evolution strategies in the context of next-generation PON, where gradual, demandbased migration demonstrates a number of significant benefits. KEYWORDS | Capacity; evolution; migration; passive optical network (PON); upgrade I. INTRODUCTION The access network is a segment of the network connecting commercial and residential subscribers to the central office (CO). Because of its impact on users’ broadband experience, Manuscript received July 23, 2011; revised October 21, 2011; accepted October 23, 2011. Date of publication January 18, 2012; date of current version April 18, 2012. G. Kramer is with Broadcom Corporation, Petaluma, CA 94954 USA (e-mail: gkramer@broadcom.com). M. De Andrade is with the Department of Electronics and Information, Politecnico di Milano, Milano 20133, Italy (e-mail: deandrade@elet.polimi.it). R. Roy was with the Department of Computer Science, University of California at Davis, Davis, CA 95616 USA. He is now with Cisco, Inc., San Jose, CA 95134 USA (e-mail: rroy@ucdavis.edu). P. Chowdhury is with the Department of Computer Science, University of California at Davis, Davis, CA 95616 USA (e-mail: pchowdhury@ucdavis.edu). Digital Object Identifier: 10.1109/JPROC.2011.2176690 1188 Proceedings of the IEEE | Vol. 100, No. 5, May 2012 the access network is considered one of the critical parts of the network hierarchy (known as the Bfirst mile[). Residential subscribers demand first-mile access solutions that provide high bandwidth and offer media-rich services. Similarly, corporate users require high-capacity and lowlatency broadband infrastructure, through which they can connect their local-area networks to the Internet backbone. Optical-fiber-based access networks have shown the ability to satisfy such consumer demands as well as a potential to accommodate future growth. While various optical access architectures, such as point-to-point (P2P) dedicated fiber or active optical networks (where aggregation switches are placed in the field) have been proposed and even tried in various deployments, passive optical network (PON) has emerged as the most flexible, scalable, and future-proof optical access technology. The flexibility of PON lies in its simple point-to-multipoint topology, low-cost implementation, and relative ease of deployment. A major factor in the success of PON is its ability to share among its subscribers the underlying network resources, such as physical fiber plant, communication channel capacity, ports at the CO, frequency spectrum, etc. By employing time division multiplexing (TDM), a single wavelength channel (and hence a single optical port) can serve many broadband subscribers in a cost-effective manner. While TDM-based PONs are most mature and are being actively deployed worldwide, other PON architectures, such as wavelength division multiplexing (WDM) PON, hybrid PON, and orthogonal frequency division multiplexing (OFDM) PON, are being explored. Recently, much attention has been given to access architectures that integrate PON with other legacy access technologies, such as Cable Modem or Wireless to create a new generation of hybrid access network architectures. 0018-9219/$31.00 Ó 2012 IEEE Kramer et al.: Evolution of Optical Access Networks: Architectures and Capacity Upgrades These new access architectures are seeking to strike the key balance between the improved performance on one side and the reduced deployment cost and minimal customer impact on the other side. However, as traffic volumes and number of heavy users continue to grow, the industry is beginning to look for the next-generation access technologies. A key requirement to any next-generation access technology is a seamless and gradual migration from the currently deployed solution to the next-generation solution. In this paper, we present the evolution of PON from two different perspectives: 1) evolution of access architectures by adopting PON as a feeder technology for various other solutions, and 2) evolution of PON itself towards a higher-capacity access. We believe that both of these evolution viewpoints must be explored to help the reader understand the issues and challenges of deploying the next generations of access networks. The rest of the paper is organized as follows. We start with an overview of existing, commercially available PON technologies in Section II. Architectural integration of PON with various nonoptical technologies is described in Section III. Section IV cites some of the principal migration requirements for PON capacity upgrades. In Section V, we discuss the candidate technologies for the next-generation PON. In order to properly evaluate various next-generation access candidate technologies, we consider not only the merits of each new technology, but also the migration scenarios from the currently deployed solutions. Finally, concluding remarks are given in Section VI. II . COMMERCIALLY AVAILABLE PON TECHNOL OGIES In recent years, PON has emerged as the most successful and widely deployed broadband technology, due to its potential to meet bandwidth demands. A PON is a point-tomultipoint (physical topology) optical network, where an optical line terminal (OLT) at the CO is connected to many optical network units (ONUs) at customer premises through one or multiple 1 : N optical splitters, as shown in Fig. 1. The network between the OLT and the ONU is passive; i.e., it does not contain active electronic devices and, therefore, does not require any power supply. A typical PON deployment uses a single fiber from the CO to the splitter and employs two or more wavelength channel(s) to communicate with ONUs. The dominant PON technologies, namely Ethernet PON (EPON, standardized by IEEE) and Gigabit PON (GPON, standardized by ITU-T), use a TDM access (TDMA) mechanism to share the communication channels among its users. Even though there are some differences in the specifications of these two technologies, their performance is largely determined by the underlying principle of TDM. TDM PON uses a single wavelength in each directionVdownstream (OLT to ONUs) and upstream (ONUs Fig. 1. A passive optical network (PON). to the OLT)Vand the wavelengths are multiplexed on the same fiber through coarse WDM (CWDM). Each ONU transmits in its assigned transmission windows, whose size and frequency are based on traffic volumes at this and other ONUs. Thus, the bandwidth available on a single wavelength is flexibly shared among all end users. Such a solution was envisaged primarily to keep the cost of the access equipment low and to make commercial deployments economically feasible. In the following sections, we discuss standardized PON technologies that have been widely deployed over the last decade and/or are in predeployment trails in various operators’ labs. A. 1G-EPON The first generation of EPON, specified by IEEE 802.3ah, provides bidirectional 1-Gb/s links using 1490-nm wavelength for downstream and 1310-nm wavelength for upstream, with 1550 nm reserved for future extensions or additional services, such as analog video broadcast. EPON uses the same MAC found in any IEEE 802.3 (Ethernet) compliant devices. The point-to-multipoint connectivity is supported by the multipoint control protocol, which uses standard Ethernet frames generated in the MAC layer. The adaptation of low-cost optics and flexible bandwidth allocation has greatly influenced the mass deployment of 1G-EPON systems. B. 10G-EPON In 2009, 10G-EPON, which is a successor of 1G-EPON, was standardized by the IEEE 802.3av task force. This technology is currently being tested by various network operators in preparation for commercial deployments. 10G-EPON supports symmetric 10-Gb/s downstream and upstream, and asymmetric 10-Gb/s downstream and 1-Gb/s upstream data rates [1]. 10G-EPON is compatible with legacy 1G-EPON and can coexist on the same fiber plant. To lower the cost of 10G-EPON implementations, a Vol. 100, No. 5, May 2012 | Proceedings of the IEEE 1189 Kramer et al.: Evolution of Optical Access Networks: Architectures and Capacity Upgrades balance between performance of optical transceivers and complexity of electronics was considered. To extend the power budget while keeping the optical transceiver parameters relaxed, the 10G-EPON specifications include a mandatory forward error correction (FEC) encoding, which also reduces deployment cost. Considering the highcapacity and low-cost implementation possibilities of 10GEPON, it could be the de facto broadband solution in foreseeable future. C. GPON GPON was developed by the ITU-T as the G.984 series of Recommendations. The focus of this work was to develop a universal PON architecture, able to deliver a mix of variable-size frames, ATM cells, and native TDM. It supports asymmetric bit rates of 2.488 Gb/s downstream and 1.244 Gb/s upstream. The GPON architecture supports a two-wavelength CWDM scheme similar to EPON. Additional downstream wavelength is allocated for distribution of analog video service. D. XG-PON XG-PON architecture has been recently standardized in ITU-T (the G.987 series of Recommendations). It supports coexistence with GPON on the same fiber plant and provides 10-Gb/s downstream and 2.5-Gb/s upstream bit rates. III . I N TE G R AT I ON OF OPT ICA L ACCESS NETWORKS Optical access networks based on TDM-PON is a success story in first-mile broadband access. Currently, there are several deployments of optical access architectures, generically known as fiber-to-the-x (FTTx); some of them are: fiber-to-the-node (FTTN), fiber-to-the-curb (FTTC), fiberto-the-premises (FTTP), fiber-to-the-home (FTTH), etc. Obviously, the deeper fiber can penetrate towards the subscribers, the more bandwidth can be provided. Even though FTTH is the most attractive technology for reasons of performance and operational simplicity, PON has been deployed in other FTTx architectures by being integrated with legacy broadband access technologies. In this section, we describe some of these developments to illustrate the architectural evolution of PON. A. PON With Copper Drops Most existing houses and apartments have some sort of copper wiring (twisted pair for phone, or coax cable for TV; most often they have both). A new wiring (fiber or copper alike) within residences is a very complicated, expensive, and complaint-prone process. Hence, network operators are interested in utilizing the copper wiring already existing in the customer premises. In this architecture, ONUs can be mounted outside residences (allowing easy access by technicians), in nearest cabinets or even on poles, and existing copper (twisted pair or coax cable) can be used as a 1190 Proceedings of the IEEE | Vol. 100, No. 5, May 2012 drop to the house and as internal wiring. One major problem of this architecture is that ONUs have to be environmentally hardened. In some cases, network operators must also supply power to the wiring closet, incurring major operational expenditures (OpEx). Two different architectures are used, depending on the type of copper cable present: 1) PON+DSL [2]; and 2) PON+MoCA (multimedia over coax alliance) [3]. In the PON+DSL topology, fiber from the CO serves a DSL access multiplexer (DSLAM) through the ONU [2]. Then, one DSLAM supports several DSL modems over a copper twisted pair links. Different DSL technologies, such as ADSL2+, VDSL, etc., can be deployed depending on reach and capacity requirements. In PON+MoCA architecture [3], coaxial cable is used to carry both data and video from the ONU using the MoCA protocol. Compared to true PON alternatives, PONs with copper drops often have bandwidth limitations due to various signal impairments in the copper cables. B. Cable Over PON Cable operators also recognized the importance of including PON-based technologies in their plant to increase capacity. This PON-based architecture allows significantly higher bandwidth than is possible with coax plants by replacing the cable modem termination system (CMTS)/modems with OLTs and ONUs. It also leverages the extensive back-office system that MSOs deployed (billing, provisioning, maintenance, diagnostics, etc.). Recently, Cable Labs proposed a supplement to the Data Over Cable Service Interface Specification (DOCSIS) that enables provisioning DOCSIS over EPON [4]. The specification, called DOCSIS provisioning over EPON (DPoE), provides a service overlay for CMTS over the EPON network. In this architecture, a CO houses the DOCSIS emulation system and the OLT. Fiber from the OLT terminates at the DPoE ONU, which, in turn, serves customer premise equipment (CPE) over coaxial cable. DPoE systems use EPON MAC and PHY layers combined with higher layers of DOCSIS protocols [4]. C. Metro–Access Integration Recently, there has been some interest in consolidating metro and access networks into a single network. The objectives of such consolidation are to lower the operator’s capital expenditures (CapEx) and OpEx by reducing the number of local exchanges and points of presence (PoPs). The long-reach access directly connected to the backbone networks has a potential to simplify the hierarchical telecom network architecture. The long-reach PON (LRPON) [5] concept is developed with such network consolidation in mind. Instead of the traditional Btree-and-branch[ topology of PON, LR-PON exhibits a Bring-and-spur[ topology where each PON segment is a Btree-and-branch[ architecture, and the OLT connects the PON segments through remote nodes connected by a fiber ring [5]. This fiber ring is leveraged from existing metro fiber ring, thereby consolidating metro and Kramer et al.: Evolution of Optical Access Networks: Architectures and Capacity Upgrades access resources into a single topology. LR-PON exhibits a hybrid TDM/WDM PON architecture, where the metro ring carries several wavelengths, and each wavelength may serve one or several TDM PON segments. LR-PON takes advantage of improved technologies for reflective semiconductor optical amplifiers (RSOAs) and burst-mode receivers to achieve the needed functionalities [5]. D. PON and Wireless Integration With the proliferation of broadband devices, such as smartphones, tablets, WiFi-enabled TV sets, etc., network operators are also looking for smart and cost-effective broadband solutions for untethered access. Wireless-optical broadband access network (WOBAN) [6] (also known as fiber-wireless (FiWi) broadband access networks [7]) is a novel network paradigm that extends the broadband access to wireless users in a cost-effective manner. WOBAN eliminates the need to deploy any cables in the residences. Hence, it is the least invasive and most cost-effective method to reach consumers. However, wireless bandwidth can be a limiting factor for some subscribers. WOBAN uses a wireless front-end (e.g., WiFi, WiMAX, LTE, etc., or a combination thereof) to connect end users, and has a wireline (e.g., PON) backhaul. In WOBAN, a traditional PON segment starts from the OLT at the telecom CO and ends at the ONUs near the wireless frontend. In a wireless mesh front-end, ONUs connect to the wireless gateways; these gateways drive several wireless routers in the wireless mesh to serve end users [6]. In case of cellular (e.g., WiMAX) front-end, base stations (BSs) are connected to the ONUs, and BSs provide untethered broadband service to the end users [7]. WOBAN not only provides a cost-effective access solution, but also exhibits several important benefits, such as service protection and energy savings. In the case of network failure (fiber cut, ONU/laser failure, etc.), WOBAN’s wireless front-end can reroute the traffic to different ONUs. This flexibility of wireless front-end can also be utilized for energy savings. During low-load hours, network operators may remotely shutdown some wireless nodes and ONUs, while ensuring that the remaining active nodes can give the necessary coverage to the remaining active users [8]. When traffic demand increases, operators may turn on some of the previously shutdown nodes to carry the traffic surge. This mode of operation increases the network utilization and saves overall network energy consumption in WOBAN. IV. CAPACI TY UPGRADE The ongoing traffic growth fueled by high-definition TV, IPTV, video-on-demand, 3-D video, wireless traffic backhauling, etc., requires a corresponding increase in the capacity of access networks. Access networks continue to be challenged to serve more users, to offer higher bandwidth, and to cover longer distances. To become successful, any access technology should satisfy a number of important operational requirements. 1) Cost efficiency refers to minimal CapEx for the given performance. 2) Resource efficiency is the maximization of resource utilization. Sharing fiber or sharing channel capacity are examples of resource efficiency. Resource efficiency often leads to cost efficiency, as fewer resources (fibers, optical ports, etc.) may be required to serve the same population of customers. 3) Scalability is a network’s ability to accommodate a greater number of users with gracefully decreasing performance (as opposed to having a hard limit on the number of users, which, if exceeded, makes performance unacceptable for all users). 4) Energy efficiency is a key characteristic. It can be static (by reducing the number of components or devices that consume more energy), or dynamic (by operating with sleeping-mode or stand-by schemes). Sharing of resources often results in improved energy efficiency. 5) Reliability is important, especially considering the trend to increase the number of users served by a single network (and are, thereby, dependent on a single point of failure). Network architectures should be able to provide protection at least to the feeder fiber and the central nodes. High levels of broadband access penetration and unprecedented rates of ongoing deployments make it highly probable that, in most areas, next-generation access networks will have to coexist with a previous-generation technology (a situation known as Brownfield deployments). Even if some deployments use a Greenfield scenario (i.e., a first deployment without any preexisting network infrastructure), network operators prefer to use a single technology for both Greenfield and Brownfield deployments. Therefore, in addition to the operational requirements listed above, the next-generation access architectures should satisfy a number of specific migration requirements. 1) Coexistence: allowing several generations of the technology to operate in the same network without affecting their service and quality. 2) As-needed/user-by-user upgrades facilitate gradual migration path, where only some of the network components are upgraded to satisfy the specific demand of some users, as opposed to a forklift upgrade of the entire network. This allows a distributed investment over longer periods of time. 3) Minimizing the disruptions: avoiding changes in the optical fiber infrastructure that can generate service disruptions. Protection schemes could reduce the effect on disruptions that can affect the entire network. 4) Reuse of outside plant: maximizing the profit of already deployed infrastructure by reusing existing Vol. 100, No. 5, May 2012 | Proceedings of the IEEE 1191 Kramer et al.: Evolution of Optical Access Networks: Architectures and Capacity Upgrades fiber, remote nodes and installed devices. Changes to outside plant in PON are usually highly disruptive to all users on that PON. Access networks should be assessed not as isolated products, but as ecosystems that enable continuous evolution. The first step in this continuous evolution was demonstrated by both IEEE and ITU-T standards bodies, which developed new standards for the second-generation TDM-PON (10G-EPON and XG-PON). These new standards focused on line-rate upgrade and are examples of capacity upgrade that satisfies all the migration requirements mentioned earlier. Both 10G-EPON and XG-PON support coexistence with the previous-generation EPON and GPON and allow a gradual upgrade. The new 10G-ONUs may be added to the network as needed (according to users’ new traffic demands) without interfering in the operation of existing ONUs. The new 10-Gb/s downstream channel uses a different wavelength channel that is only received by new 10G-ONUs. In the symmetric service, the upstream 10-Gb/s signal uses the same band as the legacy’s upstream signal in the IEEE standard (the OLT is able to receive signals at multiple line rates), or uses a different wavelength in the ITU-T standard. Disruptions to existing users are also minimized. In the case of 10G-EPON, the existing OLT line card needs to be replaced by a new (dual-rate) line card. In the case of XG-PON, a new WDM splitter may need to be installed in the CO. Or, if the splitter is integrated into a line card, a new line card should be installed instead of the old one. All these operations may be performed in a single location only (the central office) and do not take more than a few minutes, typically allocated in a maintenance window during lowest usage period. 10G-EPON and XG-PON technologies can coexist with legacy EPON or GPON and do not require any changes to the outside plant. Both technologies continue to share the capacity in time (TDM) over each wavelength, maintaining the original design objective of utilizing the network resources in a cost-effective way. While 10G-EPON and XG-PON capacity is expected to satisfy the demand for many years to come, investigation of potential candidate technologies for the third-generation of PON-based access networks has already began [9]–[12] V. CANDIDATE T ECHNOLOGIES FOR THE NE XT- GE NE RAT ION PON A number of different channel access mechanisms, such as wave division multiplexing (WDM), hybrid access (WDM+TDM), and OFDM access (OFDMA) are interesting candidates for the next-generation PON architecture. Of course, nondisruptive evolution of exiting TDM-PONs is also expected to produce higher capacity PONs. In the following sections, we briefly discuss those technologies. 1192 Proceedings of the IEEE | Vol. 100, No. 5, May 2012 A. WDM PON The WDM PON technology creates dedicated wavelength channels to achieve P2P connectivity between the OLT and individual ONUs on top of the point-to-multipoint physical topology. The technology first emerged as early as the mid-1990s and got a renewed interest in recent years, as some technological advances were reported in the literature. Several WDM PON systems are available for trials [13]. The state-of-the-art experimental WDM PON system can support 100 Mb/s–2 Gb/s symmetric communication per wavelength channel with 32 ONUs [13], [14]. WDM-PON architectures include wavelength-routed WDM PON and broadcast-and-select WDM-PON. Irrespective of the underlying technology, WDM PON provides P2P connectivity, as opposed to the point-to-multipoint principle of the TDM-PON. In wavelength-routed WDM-PON, in the downstream direction, the wavelength channels are routed from the OLT to the ONUs by a passive arrayed waveguide grating (AWG) instead of a passive power splitter used in the TDM PON (Fig. 2). The AWG, which is an optical wavelengthrouting device, routes a unique wavelength to every ONU on the same fiber plant. A generic colorless wideband (wavelength agnostic) receiver is used in the ONU to receive the ONU-specific wavelength. A multiwavelength source at the OLT is used for transmitting multiple wavelengths to the various ONUs. Among various WDM PON architectures, a solution based on the wavelengthlocked Fabry–Perot laser diodes (FP LDs; i.e., the laser excites only one mode when a well-adjusted external optical signal is coming in) has become the most popular. This architecture uses a spectrum-sliced broadband light source (BLS; originated from the OLT) to excite the FP LDs residing in the ONUs for upstream communication. When a well-adjusted external optical seed signal is applied, the laser excites only one mode corresponding to the frequency of the seed signal. A cost-efficient solution uses colorless ONUs based on RSOA [15] and sources of seed light located at the OLT. The RSOA-based ONU modulates the received seed light with the corresponding data, amplifies the optical signal, and sends it back in the upstream direction towards the OLT. RSOA-based ONUs can also be selfseeded [16]. In this case, the ONU generates its own optical signal with the help of a fiber Bragg grating (FBG) without additional light sources at the OLT. A wavelength-routed WDM PON presents a number of migration problems. It requires changes to the outside plant (to replace the power splitter with the AWG) that would result in major service disruptions. It does not allow coexistence with previous generations of deployed devices. Instead, to upgrade a user from TDM PON to WDM PON, all users will have to be upgraded at the same time. Such forklift upgrades are not practical and do not allow operators to recoup their investments, as users not interested in paying for better service or more bandwidth still get their devices upgraded. Kramer et al.: Evolution of Optical Access Networks: Architectures and Capacity Upgrades Fig. 2. Example of a WDM PON. In broadcast-and-select WDM PON, all wavelengths reach all ONUs, but each ONU is tuned to a specific wavelength [9], [17]. This architecture has a power splitter at the remote node, which is compatible with the outside plant used by currently deployed TDM-PON architectures. However, ONUs must either use a tunable laser or have a fixedwavelength laser with a unique wavelength for each PON. The fixed-wavelength lasers are cheaper than the tunable lasers, but using ONUs with different fixed-wavelength lasers entails enormous operational challenges to maintain inventories of many different types of ONUs, as well as service issues when a wrong-color ONU is connected to a PON. A study in [18] presented ultradense WDM-PON, which is a variation of broadcast-and-select WDM-PON. Ultradense WDM-PON uses large number of 3-GHzspaced channels and relies on coherent detection. This proposal has a number of advantages: the optical outside plant remains the same as was deployed for TDM-PONs. The coherent reception increases the sensitivity of the receivers, thus allowing larger distances between OLT and ONUs. However, the disadvantages associated with broadcast-and-select WDM-PONs remain. The commercialization of ultradense WDM-PON is difficult due to the complexity and cost of this technology. B. Hybrid PON While WDM PON has the potential to provide higher aggregated capacity than TDM PON, it does not necessarily guarantee better service quality or network utilization. Broadband access architectures should take into account the known user behavior. It has been reported in the literature that the behavior is nonuniform. In many instances, the usage behavior follows the Pareto principle, where 80% of the traffic is generated by only 20% of the users [19]. However, according to some network operators, the asymmetry is even higher, with almost 80%–90% of traffic being generated by only 10% of heavy users. Because of this asymmetry, WDM PON, which provides dedicated channels per users (per ONUs), is unable to share the network resources efficiently under such nonuniform load conditions. Channels that serve heavy users become saturated, while channels dedicated to light users remain underutilized. Hybrid PON attempts to combine the large aggregated capacity of WDM PON with the efficient utilization of resources of TDM PON. Hybrid PON adopts multiple channels, as in WDM PON, while allowing channel sharing among multiple ONUs, as in TDM PON, could be a balanced solution [20]. Similarly to WDM-PON, hybrid PON can be classified as wavelength-routed hybrid PON or broadcast-and-select hybrid PON. In broadcast-and-select architecture, also called stacked PON, several CWDM channels are allocated in each direction. Multiple ONUs, tuned to the same channel, operate as if they were connected to a TDM-PON. As in broadcast-and-select WDM-PON case, the ONUs require either a specific fixed-wavelength laser, a laser-array, or a tunable laser. In the case of a laser array or a tunable laser, the hybrid PON can further optimize the performance by dynamically allocating bandwidth not only in time domain, but also in time and wavelength domains. A wavelength-routed hybrid PON uses a combination of AWG and power splitters, as shown in Fig. 3. This architecture allows identical (colorless) ONUs and could potentially support more channels than the broadcast-andselect hybrid PON; however, it restricts the bandwidth to the ONUs connected to a single wavelength channel. While hybrid PONs, in general, have better operational characteristics (scalability, resource utilization, and performance) than the WDM-PONs, they suffer from the same migration challenges. Necessary changes to the outside plant, the potential need to retrofit the deployed ONUs with additional wavelength filters, and, in some Vol. 100, No. 5, May 2012 | Proceedings of the IEEE 1193 Kramer et al.: Evolution of Optical Access Networks: Architectures and Capacity Upgrades Fig. 3. Example of a hybrid TDM/WDM PON. cases, expensive or nonidentical ONUs complicate the migration from existing TDM PONs to hybrid PONs. C. OCDM PON Orthogonal code division multiplexing (OCDM) PON uses orthogonal codes to segregate different connections. The issues of noise and interference among different connections remain challenges, and the search for new codes to allow an increased number of channels continues. So far, demonstrations combining OCDMA and TDM PON make use of 2, 4, and 16 optical codes [21]. Coders/decoders and corresponding transceivers are still in the early stages of development. Few orthogonal codes can be implemented to create hybrid WDM/CDM PON [22], which would improve the scalability and flexibility of this option, at the cost of increasing the complexity of the network. Deployment of this architecture would require all end nodes to be upgraded, which is a major detriment. D. OFDM PON OFDM divides a high-bandwidth signal into many partially overlapping, yet noninterfering, lower bandwidth subcarriers. From a networking point of view, each optical OFDM subcarrier can be regarded as a transparent pipe for the delivery of arbitrary network traffic. These subcarriers can be dynamically assigned to different services and/or different users depending on the specific requirements. By subdividing an OFDM band among multiple users and/or combining OFDM with TDMA, each OFDM subcarrier can be further split among different services/users in different time slots in a dynamic fashion. As a result, in OFDMA-PON, dedicated subchannels, which can be com1194 Proceedings of the IEEE | Vol. 100, No. 5, May 2012 posed of one or more OFDM subcarriers and/or time slots, become fine-grained transparent pipes for the delivery of arbitrary analog or digital signals. In recent years, OFDMAPON has demonstrated 40- and 100-Gb/s [23] aggregated transmission rates. Since all the end-nodes need to be upgraded, the requirements of coexistence, disruption minimization, and gradual user-by-user upgrade are not satisfied. E. Higher Rate PON Constant progress in the P2P line rates makes higher rate PONs a simple and appealing solution for the nextgeneration access. The recently completed IEEE 802.3bg standard for 40-Gb/s Ethernet over single-mode fiber provides the necessary building block for the nextgeneration EPON (40G-EPON). Perhaps, as a start, an asymmetric version with 40 Gb/s downstream and 10 Gb/s upstream will be developed. As was demonstrated by new standards for the second-generation TDM-PON (10G-EPON and XG-PON) and ongoing predeployment trials of these technologies, line upgrade is an evolutionary step that satisfies all operational and migration requirements, as listed in Section IV. VI. CONCLUSION The deployed PON solutions affect how access networks will evolve and how future capacity upgrades will take place. The enormous pace of current deployments and announced plans for new deployments show that the two existing generations of PON (1G-EPON/GPON and 10G-EPON/XG-PON) will have a very large installed base before the 10-Gb/s PON capacity becomes an issue. Kramer et al.: Evolution of Optical Access Networks: Architectures and Capacity Upgrades In all areas with already existing PON deployment, the only successful next-generation architectures will be the ones that support coexistence, gradual upgrades, and minimal disruptions. Many of the major telecom operators already have or will soon have some flavor of PON-based optical access. It is highly unlikely that these operators will procure different architectures for areas with preexisting deployments and for Greenfield deployments. The most common scenario is to deploy the same access architecture in all areas. Line rate upgrades remain a most viable option. Historically, it has been the case that, as P2P Ethernet components for a certain data rate mature, EPON adopts the same data rate. That happened twice already, with 1-Gb/s EPON being standardized in 2004 (six years after the 1-Gb/s P2P Ethernet standard), and 10-Gb/s EPON REFERENCES [1] K. Tanaka, A. Agata, and Y. Horiuchi, BIEEE 802.3av 10G-EPON standardization and its research and development status,[ J. Lightw. Technol., vol. 10, no. 4, pp. 651–661, Feb. 2010. [2] G. Kramer, Ethernet Passive Optical Networks. New York: McGraw-Hill, Mar. 2005. [3] MoCA Alliance. [Online]. Available: http:// www.mocalliance.org/ [4] DOCSIS Provisioning of EPON Specifications, DPoE-SP-ARCHv1.0-I01-110225, Feb. 2011. [5] H. Song, B.-W. Kim, and B. Mukherjee, BLong-reach optical access networks: A survey of research challenges, demonstrations, and bandwidth assignment mechanisms,[ IEEE Commun. Surv. Tut., vol. 12, no. 1, pp. 112–123, Jan. 2010. [6] S. Sarkar, P. Chowdhury, S. Dixit, and B. Mukherjee, BHybrid Wireless-Optical Broadband Access Network (WOBAN),[ in Broadband Access Networks: Technologies and Deployment. New York: Springer-Verlag, 2009. [7] N. Ghazisaidi, M. Maier, and C. M. Assi, BFiber-wireless (FiWi) access networks: A survey,[ IEEE Commun. Mag., vol. 47, no. 2, pp. 160–167, Feb. 2009. [8] P. Chowdhury, M. Tornatore, S. Sarkar, and B. Mukherjee, BBuilding a green wireless-optical broadband access network (WOBAN),[ J. Lightw. Technol., vol. 28, no. 16, pp. 2219–2229, Aug. 2010. [9] J. Kani, F. Bourgard, A. Cui, A. Rafel, M. Campbell, R. Davey, and S. Rodrigues, BNext-generation PONVPart I: Technology roadmap and general requirements,[ IEEE Commun. Mag., vol. 47, no. 11, pp. 43–49, Nov. 2009. being standardized in 2009 (seven years after the 10-Gb/s P2P Ethernet standard). The natural question is whether this evolution will continue, perhaps leading to 40-Gb/s EPON that builds upon the recently completed IEEE 802.3bg standard for 40-Gb/s serial Ethernet over singlemode fiber. WDM-PON, hybrid PON, OCDMA PON, or OFDM PON technologies are not nearly as mature as TDM PON today. The longer it takes for these technologies to mature, the fewer operators will remain without any sort of legacy PON deployed. It remains to be seen whether even the dilatory operators will be willing to deploy these alternative PON technologies. The cost of equipment generally depends on volumes, and if many operators continue to deploy TDM PON, it is doubtful that alternative access technologies will get any significant foothold. h [10] M. De Andrade, G. Kramer, L. Wosinska, J. Chen, S. Sallent, and B. Mukherjee, BEvaluating strategies for evolution of passive optical networks,[ IEEE Commun. Mag., vol. 49, no. 7, pp. 176–184, Jul. 2011. [11] J. Zhang, N. Ansari, Y. Luo, F. Effenberger, and F. Ye, BNext-generation PONs: A performance investigation of candidate architectures for next-generation access stage 1,[ IEEE Commun. Mag., vol. 47, no. 8, pp. 49–57, Aug. 2009. [12] F. An, K. Kim, Y. Hsueh, M. Rogge, W. Shaw, and L. Kazovsky, BEvolution, challenges and enabling technologies for future WDM-based optical access networks,[ in Proc. 2nd Symp. Photon. Netw. Comput., Cary, NC, pp. 1–4, Sep. 2003. [13] LG Nortel, BEthernet over WDM PON technology overview,[ White paper, 2009. [Online]. Available: http://www.lg-nortel.com/ global.support.retrieveDocumentation.laf? mid=3&sid=1&item_id=10449&category=55. [14] H. Lee, H. Cho, J. Kim, and C. Lee, BA WDM PON with an 80 Gb/s capacity based on wavelength-locked Fabry-Perot laser diode,[ Opt. Exp., vol. 18, no. 17, pp. 18077–18085, Aug. 2010. [15] C. Arellano, C. Bock, J. Prat, and K. Langer, BRSOA-based optical network units for WDM-PON,[ in Proc. Opt. Fiber Commun./ Nat. Fiber Opt. Eng. (OFC/NFOEC) Conf., Mar. 2006, DOI: 10.1109/OFC.2006.215347. [16] J. Kang, S. Lee, H. Kwon, and S. Han, BOptical network unit based on self-seeded reflective semiconductor optical amplifier for WDM-PON,[ in Proc. SPIEVInt. Soc. Opt. Eng., 2006, vol. 6354, DOI: 10.1117/12.689514. [17] F. Effenberger, H. Mukai, S. Park, and T. Pfeiffer, BNext-generation PONVPart II: Candidate systems for next-generation PON,[ IEEE Commun. Mag., vol. 47, no. 11, pp. 50–57, Nov. 2009. [18] H. Rohde, S. Smolorz, E. Gottwald, and K. Kloppe, BNext generation optical access: 1 Gbit/s for everyone,[ in Proc. Eur. Conf. Opt. Commun., Vienna, Austria, pp. 3–5, Sep. 2009. [19] Sandvine, BFall 2010 Global Internet PhenomenaReport,[2010.[Online].Available: http://www.sandvine.com/downloads/ documents/2010%20Global%20Internet% 20Phenomena%20Report.pdf. [20] M. De Andrade, M. Tornatore, S. Sallent, and B. Mukherjee, BOptimizing the migration to future-generation Passive Optical Networks (PON),[ IEEE Syst. J., vol. 4, Special Issue on Broadband Access Networks, no. 4, pp. 413–423, Dec. 2010. [21] H. Iwamura, H. Tsuji, H. Tamai, M. Sarashina, N. Minato, M. Kashima, and T. Kamijoh, BA study of 160 Gbps PON system using OTDM and OCDM technologies,[ in Proc. Opt. Fiber Commun./Nat. Fiber Opt. Eng. (OFC/NFOEC) Conf., Feb. 2008, DOI: 10.1109/OFC.2008.4528148. [22] K. Kitayama, X. Wang, and N. Wada, BOCDMA over WDM PON-solution path to gigabit-symmetric FTTH,[ J. Lightw. Technol., vol. 24, no. 4, pp. 1654–1662, Apr. 2006. [23] N. Cvijetic, D. Qian, and J. Hu, B100 Gb/s optical access based on optical orthogonal frequency division multiplexing,[ IEEE Commun. Mag., vol. 48, no. 7, pp. 70–77, Jul. 2010. ABOUT THE AUTHORS Glen Kramer received the M.S. and Ph.D. degrees in computer science from the University of California at Davis, Davis, in 2000 and 2003, respectively, where he was awarded a National Science Foundation (NSF) grant to study nextgeneration broadband access networks. He is a Technical Director of Ethernet Access at Broadcom Corporation, Petaluma, CA. He has joined Broadcom through its acquisition of Teknovus, Inc., where he served as Chief Scientist. He has done extensive research in areas of traffic management, quality of service, and fairness in access networks. Prior to Teknovus, he worked at the Advanced Technology Lab at Alloptic, Inc., where he was responsible for design and performance analysis of PON scheduling protocols and was involved in prototyping the very first EPON system. He authored 16 patents. His book Ethernet Passive Optical Networks has been published in English (New York, NY: McGraw-Hill, 2005) and Chinese (Beijing, China: BUPT Press, 2007). Dr. Kramer chairs the IEEE P1904.1 Working Group that develops a standard for Service Interoperability in Ethernet Passive Optical Networks. Previously he served as chair of IEEE P802.3av B10 Gb/s Ethernet Passive Optical Networks[ task force and as EPON protocol clause editor in IEEE 802.3ah BEthernet in the First Mile[ task force. Vol. 100, No. 5, May 2012 | Proceedings of the IEEE 1195 Kramer et al.: Evolution of Optical Access Networks: Architectures and Capacity Upgrades Marilet De Andrade received the Electronic Engineering degree from Simon Bolivar University, Miranda State, Venezuela, in 1998, the M.S. degree in systems and communication networks from the Polytechnic University of Madrid, Madrid, Spain, in 2003, and the Ph.D. degree in telematics engineering from Polytechnic University of Catalonia (UPC), Catalonia, Spain, in 2010. Her major field of study is broadband access networks and optical communications. From 1998 to 2001, she served as Interconnections Engineer at Movistar Telefonica Venezuela (former Telcel Bellsouth). From 2003 to 2007, she held a Research Fellowship from the Spanish Ministry of Education, and worked for the Broadband Networks group at UPC. She served as Lecturer at UPC from 2007 to 2008. She spent one year (2009) at the Computer Science Department, University of California at Davis, Davis, as a visiting Ph.D. student. In 2010 and 2011, she was a Researcher at the Next Generation Optical Networks (NEGONET) group, Royal Institute of Technology (KTH), Sweden. Currently, she is a Postdoctoral Researcher at the Department of Electronics and Information, Politecnico di Milano, Milano, Italy. Her research interests are PON evolution and resource allocation in broadband access networks. 1196 Proceedings of the IEEE | Vol. 100, No. 5, May 2012 Rajesh Roy received the B.S. degree in computer science and engineering from Javadpur University, Kolkata, India, in 2006, and the M.S. and Ph.D. degrees in computer science from the University of California at Davis, Davis, in 2007 and 2010, respectively. He is currently a Software Engineer at Cisco, Inc. in the data center switching and cloud computing division. His research interests include data center switching, broadband access networks, optical WDM network upgrading, and resource optimization. Pulak Chowdhury received the B.Sc.Eng. degree from Bangladesh University of Engineering and Technology, Dhaka, Bangladesh, in 2002, the M.A.Sc. degree from McMaster University, Hamilton, ON, Canada, in 2005, and the Ph.D. degree in computer science from the University of California at Davis, Davis, in 2011. His research interests cover a variety of topics in optical networks, hybrid wireless-optical networks, and energy efficiency in next-generation networks.