Access Architecture Definition Document

Transcription

1 FP Deliverable D22 Access Architecture Definition Document June 2010 ReDeSign Research for Development of Future Interactive Generations of Hybrid Fiber Coax Networks Information for Publication: Version: 1.0 Status: Public (PU)

2 Executive summary This deliverable, called Access Architecture Definition Document, makes a summary of the research results that have been achieved in the ReDeSign project on the next generation HFC research. During the 2,5 year project all aspects of an evolution of the cable network has been analyzed and reported in different deliverables. First the requirements of the evolution of the HFC network, called Next Generation HFC, have been defined based on a questionnaire that has been answered by a lot of European cable operators and by interaction with cable operators on MSO forum meetings. A cable network makes use of two different transmission media, fiber and coax. For both technologies a detailed analysis has been performed during the project to define the most optimal technology for the realization of the next generation HFC network. Possible topologies for the creation of the NG-HFC network have been studied and discussed. In case of a shared medium, such as an HFC network, also the MAC protocol is rather important to enable efficient transmission and utilization of the network. Different MAC protocols used in the world to manage shared medium networks have been analyzed and compared. Operators do not only consider technology and performance for choosing a migration path towards the future, but also cost is a driving element. Within this project a cost analysis has been performed on existing HFC networks, and compared this with existing GPON FttH technology. In this deliverable all these aspects are summarized, for a more detailed analysis the reader is referred to previous deliverables of the project. Based on all the work that has been performed in the context of the ReDeSign project on evolutions of the HFC network, there were guidelines developed that can be given to cable operators. Assessments have proven that there is a need for the definition of a Next Generation HFC network. Upgrading the current HFC network with techniques as node splitting, analogue switch off or other proposed and used techniques will prolong the lifetime of the current HFC network, but will not take away the need for a next generation of the HFC network. The proposed solution for the realization of a NG-HFC is different in case of brownfield deployments or in case of greenfield deployments. For the greenfield deployments there is no question that Fiber to the Home (FttH) or the Building (FttB) is the optimal way forward, both on the short and long term. Activation of the fiber could happen with RFoG or TDM-PON (EPON, GPON) technologies. In a brownfield the situation is different, the deployed coax is such a big asset that it should be leveraged as long as possible. The potential capacity of the coax is also of an order that from a physical layer point of view it can compete with fiber. The point is that the available bandwidth over this coax network should be activated in the most cost and power efficient way, taking also into account possible density issues in the hub or Central Office. In this deliverables three options have been put forward, all based on bringing fiber closer to the user, combined with coax transmission for the last/first part of the network. For European cable operators, which have in most cases already a rather good bidirectional triple play ready HFC network in place, an evolution of the current DOCSIS approach is most attractive. The architecture that is put forward in Europe for the realization of the NG-HFC network is based in a network element called µ-cmts combined with a deep fiber topology. This is a new network element, located at the position of the fiber node, deep in the network, that takes care of the full coax transmission between this node and the cable modem at home. The proposed solution is fully compatible with currently used analogue and digital TV distribution concepts and technologies, and solves at the same time the density issues in the hub or other locations, while increasing strongly the potential bandwidth for the user. Transmission between the Central Office/hub and the µ-cmts is standard Ethernet/IP transmission. In order to make this solution the most cost and power efficient solution possible, a new evolution of DOCSIS will be needed. Within this deliverable some suggestions for DOCSIS standardization are given. A proposal based on the Ethernet concept, only supporting essential features and making nice-to-have features optional, would downscale the complexity of DOCSIS. 2 Access Architecture Definition Document

3 In summary a proposal for the realization of a NG-HFC network has been described in this deliverable, taking into account both brownfield and greenfield deployments. Necessary technologies have been analyzed and described. Migration paths towards this NG-HFC network are indicated. A final remark that should be given is that the technologies developed within the ReDeSign project have a place in the migration towards or in the definition of the NG-HFC network. The new standard for coax transmission, DVB-C2, will increase strongly the capacity of the coax network. In brownfield deployments still coax will be used as transmission medium in the NG- HFC network, so DVB-C2 is the best candidate for the physical layer of the coax transmission of the NG-HFC network, certainly in case DOCSIS or an evolution of DOCSIS is still used as MAC layer. The new amplifier technology as demonstrated in the project has a higher output power over a large band (+3dB till 1GHz). One of the possible migration paths between an HFC network and a NG-HFC network is the increase of the downstream frequency band till 1GHz. Practical problems with respacing of amplifiers makes this upgrade technique in reality not very attractive. The introduction of this new type of amplifier can prevent the need for respacing, making this option of upgrading the HFC network much more attractive. 3 Access Architecture Definition Document

4 Contents Executive summary...2 Contents...4 List of figures...6 List of tables Introduction Requirements of the NG-HFC network Existing HFC networks The Coax Network Topology Technology The Fiber network Topology Technology EPON WDM-PON RFoG or RF-PON Next Generation HFC networks Upgrades from existing HFC techniques More disruptive scenario s NG-HFC MAC layer functionality Traffic Management Privacy and security Digital efficiency OSS services General Considerations Greenfield networks The optimal NG-HFC network NG-HFC topologies NG-HFC technologies Fiber technology Coax technology MAC technology Cost evaluation Cost comparison of HFC and FttH Cost of FttLA Proposals for NG-HFC architectures Access Architecture Definition Document

5 5.4.1 Brownfield Greenfield Conclusions...35 Annex 1: List of abbreviations...37 Annex 2: References Access Architecture Definition Document

6 List of figures figure 1: generic figure of a HFC network...8 figure 2: Tree and Branch topology...10 figure 3: Hybrid topology...10 figure 4: Star topology...10 figure 5: Three possible FttH topologies...14 figure 6: Topology of a Next Generation HFC network...17 figure 7: Three most probable architectures for the realization of the NG-HFC network figure 8: Total investment versus time for an HFC network and a GPON FttH network, with average cost for digging included...31 figure 9: Total investment versus time for an HFC network and a GPON FttH network, digging cost not included...31 figure 10: Topologies for a FttLA approach: high density...32 figure 11: Topologies for a FttLA approach: low to medium density...33 figure 12: Different FttLA topologies as used for the cost calculations...34 List of tables table 1: Design parameters and principles currently deployed by operators...9 table 2: European coaxial network parameters, as been reported in deliverable D table 3: Comparison of different upgrade technologies...26 table 4: Comparison of the different MAC layers with the NG-HFC MAC table 5: Results of a cost analysis of the FttLA approach Access Architecture Definition Document

7 1 Introduction This deliverable D22 summarizes the work carried out by work package 6 of the project. The focus of WP6 is on the definition of a next generation HFC network, that should support cable operators to be competitive also in the future when new competition is coming from operators utilizing fiber networks. As explained in other deliverables and demonstrated by the ReDeSign project, the cable network, including the coax segments, have a large potential bandwidth. By defining new physical layer technologies the net bandwidth available to the cable operator can still strongly be increased. The project has demonstrated this by its collaboration in the DVB-C2 standardization process, improving the bandwidth over a coax segment considerably. The main question is how the network should evolve and should look like in the future in order to activate the bandwidth in the most cost, power and density efficient way. Therefore the ReDeSign project has analyzed the technologies used in HFC networks (fiber technology and coax technology) and reported the results of that study in previous deliverables. Topologies for a future evolution of the cable network have been defined, this network was called Next Generation HFC network or short NG-HFC network. In order to make optimal use of the cable infrastructure not only the physical layer technologies (coax, fiber) are important but also the method to activate the technologies, the way information is transported over the network. This is organized by the MAC-layer of the network. For digital interactive services this MAC- layer in HFC networks is currently DOCSIS. The project alsohas studied on the functionalities in the MAC-layer, different existing MAC layers have been compared and guidelines for the evolution towards the NG-MAC layer used in NG-HFC networks have been formulated. In this deliverable all above mentioned aspects, equally important for the definition of an efficient NG-HFC network, are brought together. First the requirements for a Next Generation HFC network are recapitulated. These requirements are coming from cable operators and are obtained by the project by sending out a questionnaire during the first year of the project. The results of this questionnaire have been discussed in detail in deliverable D03 and D04. Out of these results the requirements have been extracted, as reported in deliverable D09. A summary of these conclusions is given in this deliverable. The existing HFC networks are the main asset of a cable operator. The ReDeSign project has studied the European cable networks and has described these networks and its communalities and differences at length in deliverable D06. Two types of transmission techniques are used for the creation of an HFC network, coax and fiber. Within the project both techniques have been analyzed, and all different existing technologies have been compared with each other. In chapter 3 a summary is presented on the existing HFC networks, as on the coax and fiber transmission state-of-the-art technologies. A next generation HFC network should be a natural evolution of an existing HFC network, still based on fiber and coax, but should also embark on the technical evolution of the technologies in order to make it future proof. Within chapter 4 all aspects of the NG-HFC network are summarized, based on detailed studies performed during the project and reported in previous deliverables. Both topology and technology aspects are mentioned. Fiber technology as coax technology but also possible MAC technologies have been analyzed. Based on all the knowledge build up during the project lifetime clear guidelines can be formulated on how a European cable operator could evolve his network towards a NG-HFC network. Chapter 5 will give guidelines based on the analysis reported on in chapter 4, for the different aspects of the NG-HFC network. A preferred way forward in brownfield and greenfield deployments is described, based on availability and capability of technology but also from a cost analysis point of view. Finally conclusions are formulated in chapter 6. 7 Access Architecture Definition Document

8 2 Requirements of the NG-HFC network The requirements for a next generation HFC network have been analyzed based on the input from cable operators that has been received via the questionnaire, as created and distributed by the ReDeSign project. The questionnaire and its results has been described in earlier deliverables of the project (see deliverable D03, D04 and D07). In D09 this input is translated into requirements for the NG-HFC network. In this deliverable only the conclusions are listed, for more information the reader is referred to the previous deliverables. Requirements for a future network will come from different angles. Part of the requirements comes from the services an MSO wants to offer to his customers. A detailed analysis has been made for this in deliverable D07. Second part of the requirements for a next generation network comes from competition. Any network that has to be defined needs to be better than that of the competition. In case a MSO can offer all services a subscriber wants, but a competitor can offer more, subscribers will move to that service provider. As such the next generation HFC network has to beat the offering of competitors network by that timeframe. Finally MSOs will not start from scratch building a next generation network. The existing network should be leveraged as much as possible and a clear migration path from the existing network towards the next generation HFC network needs to be defined. As such this also put some requirements on the future HFC network. In order to facilitate this work, reference architectures summarizing the existing variations in networks as documented in deliverable D06 will be used. As discussed above the requirements for a next generation network come from different angles. To conclude the requirements for a next generation HFC network can be summarize as follows: At first the bandwidth offered by the network to any user should be increased, both offering more unicast interactive services and also more broadcast digital services. Analogue distribution of radio and TV will remain important but will occupy less total bandwidth of the available spectrum in the future. In the future it is expected that not only the total raw bandwidth offering will be important, but also the QoS in order to bring services with a high quality of experience. The second important set of requirements has to do with cost. The future network should realize a competitive offering at a very low cost. As such leveraging from technology also used in other areas is a plus Finally the future network should not fully break with the existing HFC concepts and network, but should leverage as much as possible the investments already made, and define clear migration paths. 3 Existing HFC networks figure 1: generic figure of a HFC network 8 Access Architecture Definition Document

9 figure 1 shows the generic figure of a HFC network. Basically an HFC network exists of a fiber part and an coax part with an optical node between the physical networks. In deliverable D06 of the ReDeSign project a study has been made for the dimensions of the European HFC networks. A part of this result is shown in table 1. Parameter Average value Remarks Homes Passed per Head End 239 k 2 groups can be distinguished, a group with relative on average small amount (between 10k and 100k) and a Homes Passed per Head End maximum Homes Passed per primary hub on average Homes Passed per primary hub maximum Homes Passed per optical node on average Homes Passed per optical node maximum Optical node served from a primary hub by the same fiber on average Homes Passed per first coaxial distribution point on average Homes passed per end amplifier on average group with a large amount (100k 1000k) 327 k 2 groups can be distinguished, one between [10k 300k] and one between [300k 1000k], on average the maximum is around 2 to 3 times the average value, but exceptions exist in both direction (max = average, 1 head-end for the whole network, 10, lot of different networks distributed over the network) 37k Most of the operators are in the range of [10k- 50k] 103 k Most of the operators give values between [10k 100k], but some give figures above 250k, impacting heavily the average value Again 2 groups can be distinguished, one between 100 and 1000 and one between 1000 and 3000, on average 2300 Most of the values are between [ ] 20 For most cases the maximum value is between [2-5] times the average value, with a few exceptions in both directions The maximum varies between [1,3 3] times the average value table 1: Design parameters and principles currently deployed by operators The fiber part of the network exists of typical network elements that should be taken into account while designing the NG. The Head End (HE) or Central Office (CO): this is the location were the cable operator has installed the main equipment for delivering the services. For instance will this be the location were the TV signals from satellite enters the network of the operator. In a centralized DOC- SIS network this will be the location were the CMTSs are placed. Distinguished characteristics of a HE is that the amount of Homes Passed (HP) it serves are very large, on average within Europe Homes Passed, see also table 1. The Local Centre (LC) is often a smaller subset of the HE and bigger than the hub. A Local Centre doesn t have to be within the network of the cable operator it is an optional location. For the reference network it will not be taken into account, but for an operator it is a possible location to place equipment for the NG-HFC network. The hub is the next segmentation of the network. In General a hub is an acclimatized location it services on average Homes Passed. Same as with the Local Centre, a Hub is not general designed for all networks. 9 Access Architecture Definition Document

10 General remark, in this deliverable CO and hub will both be used but can be interchanged. While in reality in most cases a hub will be smaller than a CO, this might also be different. Both are acclimatized locations in the network. The last element is typically for a HFC network, it is the optical node, which is the demarcation point in the network between fiber and coax. This is the point where the information from a optical medium translates into a electrical medium (and vice versa). Typically the optical node is installed in non-acclimatized street cabinets. On average it services around 1074 HP, see table 1. From the optical node towards the customer the HFC network is existing of a coax part with electrical amplifiers. Most coax networks exist of one or more group-amplifiers and at the end of the coax line, before the multitap an end-amplifier. Again the network is segmented more and more, see table 1 for the typical amounts of Homes Passed that a element services. More details about the coax network will be described in paragraph The Coax Network As explained in the introduction the HFC networks exists besides a fiber part also of a coax network. When looking at the current designs, as being used by the European cable operators, see table 1 and table 2, it is obvious that a cable operator has a large amount of coax and allocated equipment in it s network. Therefore a upgrade from an existing network to the NG-HFC network to comply to the future demands of bandwidth and associated services will be the most challenging job for an operator. Before designing the NG-HFC network, the Re- DeSign project had to study the average European network to define it s reference model on which a NG-HFC network could be build. Based on this study, see Deliverable D06, The following paragraph will shortly describe the existing coax topologies and the technologies that currently are available for use in a coax environment. figure 2: Tree and Branch topology figure 3: Hybrid topology figure 4: Star topology 10 Access Architecture Definition Document

11 3.1.1 Topology As shown in deliverable D06, the coax part of the network has basically three varieties within the European cable networks. It can be a tree and branch, see figure 2, a star, see figure 4 or even a hybrid topology, see figure 3. Not only the type of topology and the length between the branches varies but also the way that the coax networks are build even when they share the same topology, see table 2. E.g. the amount of cascades amplifiers vary from 1 till even networks with 15 amplifiers in a branch! But there are, for this deliverable, a couple of important characteristics of the coax network. The most important similarity is that in all networks coax is the transport medium that connects the customer to the cable operators network. In principle the capacity of a single coax cable is unlimited but the attenuation due to the skin-effect will limited it s use in relation to it s maximum reach. The higher the frequency of the used signal the higher the attenuation of the signal will be and thus limits it s maximum distance. Until now all networks and most of the associated electrical equipment are designed on a maximum frequency use of 862MHz. Based on this maximum frequency, the type of coax cable with its associated attenuation and the delivery of broadcast signals like analogue TV and radio the original HFC networks were designed. Originally the HFC network was designed for broadcast services like TV and radio distribution. About 15 years back operators started to roll out interactive services for which they needed a return path. Because of all legacy equipment and services operators decided to use the frequencies below 87MHz and thus the return channels or upstream band was created, nowadays often from 20 till 60MHz. Summarizing can be stated that the coax has in principle unlimited capacity, but that the capacity is limited through: o The use of limited frequency band for the upstream and downstream o The attenuation per length of the coax cable Parameter Tree-and-branch Hybrid Star Occurrence 47% 29% 18% # Cascaded distribution amplifiers on average [1-15] [2] [1-2] # Cascaded distribution amplifiers maximum [1-40]] [3-4] [2-4] # Trunks per distribution amplifier on average [1-4] [2-3] [2-12] # Trunks per distribution amplifier maximum [1-4] [2-4] [3-18] Distribution branch length on average [0,1-5] km [0,3-13] km [0,2-3] km Distribution branch length maximum [0,4-28] km [0,5-16] km [0,4-5] km # Cascaded trunk amplifiers on average [4-15] [2-4] # Cascaded trunk amplifiers maximum [3-40] [4-11] # End branches per trunk amplifier on average [2-4] [2-3] # End branches per trunk amplifier maximum [2-4] [2-4] Trunk length on average [0,2-1,5] km [0,6-8] km Trunk length maximum [0,5 12] km [1,2-12] km # End amplifiers per end branch on average [1-6] [3-4] [2] # End amplifiers per end branch maximum [2-39] [4-10] [1-3] End branch length on average [0,1-1,5] km [0,1-0,4] km [0,2-3] km End branch length maximum [0,1-1] km [0,2-0,6] km [0,3-5] km table 2: European coaxial network parameters, as been reported in deliverable D06 11 Access Architecture Definition Document

12 3.1.2 Technology Besides the coax network itself, the used techniques also have to be analyzed, to determine the migration path to the NG-HFC network. Basically all coax transmission technologies can be divided into two basic technologies: First are the analogue transmission technologies, these are legacy techniques, like FM, PAL, SECAM, used to deliver radio and television signals. Within a standard 8MHz bandwidth channel only one analogue TV-station can be broadcasted to the customers. Second are the digital transmission technologies, these techniques are mostly based on Quadrature Amplitude Modulation (QAM) and are used for transmitting digital information. There are currently two basic transmission techniques used for broadcasting digital signals: o DVB-C, used for transmitting digitalized TV stations over cable networks. The digital information is MPEG-2 encoded. The used modulation order within DVB-C are 64QAM and 256QAM which offers a total 38 to 51Mb/s bandwidth within a 8MHz channel. DVB-C is a broadcast (downstream) technique and doesn t support bidirectional communication. o EuroDOCSIS, is used for transmitting data bidirectional, and used for Internet and telephony services. The Euro in EuroDOCSIS stands for a European version of DOC- SIS. The main difference is the adaptation of European channel widths (8MHz) and spectrum use. EuroDOCSIS exists of a several versions, 1.0, 1.1, 2.0 and 3.0. From a transmission point of view 2.0 and 3.0 are very important. DOCSIS2.0 offered higher modulation, thus more capacity, in the upstream, towards the Central Office. While DOCSIS3.0 supports channel bonding to offer higher download and upload speeds. In the downstream, towards the subscribers, there is no difference between all the DOCSIS versions. the maximum modulation rate for all versions is 256QAM and offers a 51Mb/s capacity in a 8MHz channel. DOCSIS also uses MPEG-2 transport framing to transmit the data. But there are also new digital techniques, these are not (yet) implemented in a coax access network: o DVB-C2, as a part of the ReDeSign project in Work Package 5 a new transmission technology has be developed for transmitting digital information over the cable network: DVB-C2 has the possibility to use OFDM to increase the spectral efficiency. Also due to the use of new FEC coding schemes, DVB-C2 allows higher modulation rates like 1024QAM and 4096QAM which offers within a 8MHz Channel a bit rate of 70 resp. 84Mbps. Ethernet over Coax (EoC) is a general name used to describe all techniques that can be used for transmission of information over a coax access networks, different from DOCSIS. The names come from the fact that all these techniques have in common that the layer 2 protocol that is supported is Ethernet. The physical transmission of the information over the coax is different for the different approaches. In the previous deliverable a number of techniques have been analyzed, that are used in the home environment for transmission of information over coax networks, but the reader should be aware that the Ethernet over Coax family is broader than only the techniques mentioned above (MoCA, G.hn, Home-PNA). Different vendors bring solutions on the market that realize coax transmission based on a lot of concepts, e.g. WiFi over coax, ultrawideband over coax, powerline over coax are other examples of this group. It has to be mentioned that none of these solutions have been standardized yet for access networks, which does not mean that this is only theoretical. A number of propriety solutions are available on the market and some of them are being trialed by cable operators in the field. A number of solutions, based on existing standards, are listed here: 12 Access Architecture Definition Document

13 o Standard baseband IEEE802.3 Ethernet, at this moment for coax there are two standards defined 10Base2 and 10Base5 both with a maximum speed of 10Mbps. These do not meet the current demand of at least of 100Mb/s per customer. There is no standardization yet for baseband Ethernet 100Mb/s or 1Gb/s over coax. o MoCA, Multimedia over Coax Alliance, is an open standard for delivering data over the unused spectrum of coax cable. MoCA uses a 50MHz channel above 862MHz and has a maximum throughput up to 270Mb/s. MoCA can coexist with cable TV, DVB-C and DOCSIS. The frequency range within MoCA operates is MHz o ITU-T G.hn / HomeGrid, the ITU-T G.hn or HomeGrid is a standard being developed by the ITU-T. The G.hn standard is based on the MoCa technology. The main difference is that besides 50MHz bandwidth it also supports 100MHz bandwidth and allows a theoretical 762Mb/s Ethernet throughput. G.hn can operate in the frequency range from MHz. Because G.hn / HomeGrid is partly based on MoCA it can coexist with MoCa on the same medium. ITU-T G.hn / HomeGrid is standardized, while MoCA is based on corporation of members of the MoCA alliance. o HomePNA 3.1, HomePNA 3.1 uses the frequency range of 4 to 52MHz and can overcome greater distances due to low loss caused by the skin effect at high frequencies. Home PNA was originally development for delivering services over copper / telephony cables, version 3.1 includes also coax as a transmission medium. HomePNA 3.1 also includes guaranteed QoS. HomePNA is also standardized by the ITU-T. The theoretical throughput is up to 320Mb/s. Because of the used spectrum, HomePNA cannot coexist with (Euro)DOCSIS on the same coax network. Because of the limited maximum speed of 10Mb/s Ethernet is supposed not to be a suitable solution for use in the access network of the cable operator. The other three above mentioned home solutions have possibilities for use in the HFC network. HomePNA is on a small scale already been used in Hotel and Fiber to the Building solutions. On the other hand MoCA is used in the field for in-home transmission. The G.hn / HomeGrid solutions looks like the most universal and standardized solution and has the highest theoretical throughput which is from a cable operator point of view more interesting. To use one of the solutions in the HFC network the expectation will be that the solution needs deeper fiber in the HFC network, until the end amplifier. At this location a Multi Dwelling Unit (MDU) is needed to offer the baseband Ethernet, MoCA, G.hn or HomePNA. If one of the standards can be redesigned to serve more devices (2000HP according to the reference HFC network, but with node split this will be reduced to 600 or 256 or even less in a FttLA approach) and offering more theoretical capacity it should be even theoretical possible to place this MDU in a optical node. 3.2 The Fiber network The fiber part of the HFC network is the part between the Central Office or Hub and the optical node. In the optical node the optical transmission is terminated and coax transmission is started. The fiber network of existing HFC networks is mainly based on a point to point topology, sometimes a point to multipoint topology from a fiber point of view is used, but in that case each optical node has its own set of wavelengths over the fiber, implementing a point to point architecture between the hub and the optical node. Node segmentation increases the demand on the fiber part of the network. This increase is solved by adding more fibers in the network or by using WDM technology over the fiber. In case of WDM technology more than one wavelength is used for transmission between hub and fiber node (e.g. one per segment). 13 Access Architecture Definition Document

14 In case an operator is planning to extend his fiber network (e.g. for the implementation of a deep fiber architecture) he can choose from three basic topologies and associated technologies. This section will summarize the topologies and technologies that can be used for the realization of the fiber part of the HFC and next generation HFC network. A Fiber to the Home (FttH) terminology is used in this section, but the topologies and technologies are also applicable for other architectures such as fiber to the node (FttN), fiber to the last amplifier (FttLA) or fiber to the building (FttB). For more information on this topic the reader is referred to deliverable D12, here a summary of the conclusions is given Topology Basically there are three optical solutions that are generally used when deploying fiber in a greenfield FttH situation, see figure 5. Central Office Access loop Home Active Ethernet IP Ethernet switch Ethernet switch Point -to- Point IP More concentrated More distributed Ethernet switch Splicing PON IP Optical splitter PON OLT figure 5: Three possible FttH topologies 1. Active Ethernet (AE) is an Ethernet FttH architecture employing a fiber-feeding approach that supports larger deployments than P2P. AE offers long reach to a high number of customers from one centralized local exchange. With AE, fewer fibers pass through smaller ducts with greater consolidation in the CO. Its topology is similar to that of FttN; fiber extends to an aggregation node (e.g. the fiber node in an HFC network). On the downside, operational expenditures (OPEX) are typically high given the active equipment that sits vulnerable in the OSP, requiring care and maintenance. 2. Point-to-Point (P2P) is an Ethernet FttH architecture similar in structure to a twistedpair cable phone network; a separate, dedicated fiber for each home exists in the CO. On the positive side, this results in a completely passive OSP and a solution based on mainstream Ethernet technology. Weaknesses are related mainly to the massive amount of fibers in the feeder section, at least one fiber for each subscriber and the large CO, with high power consumption. 14 Access Architecture Definition Document

15 3. PON is a fully optical FttH architecture option that offers the best of all worlds. Like AE, it aggregates users in the OSP, resulting in a CO without a mess of fibers; like P2P, it employs a passive OSP device (the optical splitter). The splitter is extremely stable (it virtually never fails) and requires no cooling or powering. PON offers long reach and supports deployments on any scale. It is highly reliable and requires minimal maintenance, offering an important OPEX advantage. For these reasons, it is best suited to mass-market deployments. When assessing the merits of the various architectures for FttH deployment, scalability requirements are perhaps the most important consideration. AE, P2P and PON all differ in: the amount of fiber required the amount of CO-based hardware to be deployed the amount of power consumed at the CO/Hub, Access Loop and Home reach from CO passive or active OSP None of the technologies is better than the other in all cases, so careful analysis has to be made case by case which of the technologies to deploy Technology See deliverable D12 for a more detailed description of the used fiber technologies described below. A summary of the technologies most attractive for cable operators is given here Active Ethernet Active Ethernet is, as the name says, based on Ethernet switching techniques. In the access loop an aggregation switch is placed. All subscribers are point-to-point connected to this switch and there connection is being trunked to the main switch in the CO. The technique is based on the Ethernet standards. The maximum distances and data capacity for AE is depending on the availability of the standard Ethernet products. This makes this technique rather cheap and easy to implement, compared to PON techniques. The main drawback is that the operator has active equipment in the field that consumes energy and is more service depended than an optical fiber or a passive splitter / combiner. An Active Ethernet technology is less suited for the broadcast transmission of analogue or digital television to all users on a separate wavelength. This drawback makes this technology less attractive for cable operators GPON Gigabit PON (GPON) was defined by the Full Services Access Network (FSAN), a group of operators and specified in ITU-T Recommendation G.984. GPON technology supports integrated audio, data and video services, providing 2,5Gb/s-1,25Gb/s to/from a maximum of 64 users. GPON provides a reach of 20km with a 28dB optical budget, which can be extended to 30km if the splitting ratio is limited to 1:16. Typically, an OLT-system can cope with up to approximately 3600 subscribers, based on a maximum of 64 subscribers per GPON connection. 3 wavelengths are defined on the PON system, 1 wavelength for DS digital transmission (1490nm), 1 wavelength for US digital transmission (1310nm) and 1 wavelength for DS broadcast RF transmission (1550nm). GPON uses a Time Division Multiplexing (TDM) based scheme. In the downstream direction, data are sent to all ONTs connected with the PON OLT-interface. In the ONT, only data destined to that ONT are filtered out based on the portid sent along with the data. In the downstream direction the MAC protocol controls the assignment of transmission windows for each ONT. In this way also bandwidth allocation and prioritization is performed. In the upstream 15 Access Architecture Definition Document

16 TDMA is used as scheme to share the bandwidth between the subscribers. The scheme is rather comparable to the DOCSIS methodology; a central manager allocates the bandwidth over the users and distribute grants to the user to indicate which user can transmit traffic in which timeslot. The length of the timeslots allocated to a certain user is variable EPON Ethernet PON (EPON) is a PON technology developed by the IEEE and specified in IEEE802.3ah. It provides 1Gb/s symmetrical to/from users (proprietary 2,5Gb/s solutions are available on the market) and a reach of 20km with a 29dB optical budget. EPON uses the same fiber for both up- and downstream traffic with different wavelengths: 1490nm for downstream and 1310nm for upstream. The Multi-Point Control Protocol (MPCP) performs the bandwidth assignment, bandwidth polling, auto-discovery and ranging. Like GPON, EPON uses a TDM based scheme. Frames that are broadcasted downstream are extracted in the Optical Network Unit (ONU) (similar to the ONT of GPON) based on the logical link ID in the preamble. Upstream bandwidth is assigned by means of GATE messages sent by the OLT WDM-PON In a pure Wavelength Division Multiplexing PON separate wavelengths run over the same shared fiber infrastructure, where in the last fiber part each wavelength is directed to the right per home subscriber fiber. Dense Wavelength Division Multiplexing (DWDM) requires accurate lasers with little mode partition noise (MPN). These types of lasers are 25 to 50 times more expensive than the simpler laser types. Two of these lasers are needed per customer for PON applications, making optical components a significant fraction of the overall capital expense. o CWDM, the relatively small numbers of available CWDM channels (a maximum of 5 downstream/upstream channels) in the band of interest, together with the fact that CWDM is not compatible with TDM (difference in wavelength range used), makes CWDM unattractive for WDM PON deployments. o DWDM, in contrast with CWDM a relatively large number of channels can be packed into the allocated bands using DWDM. But this density comes with the increased cost associated with tight manufacturing tolerances of lasers as well as with stabilizing the laser within the narrower channel width. To reduce costs wider channel spacing can be employed (3.2 or 1.6nm) which can still result in many useable channels per 20nm window. However the costs of these optics is still substantially higher than for GPON optics RFoG or RF-PON Radio Frequency over Glass (RFoG) or Radio Frequency PON (RF-PON) is a technology that enables transmission of radio signals over fiber. Since the radio frequencies are the same as used on HFC networks, there is a high degree of compatibility between an RFoG system and equipment already deployed in the field by cable operators, such as the CMTS, OSS/BSS systems and CPE equipment. To interwork with an existing cable modem only a relatively simple Optical-Electrical converter is needed between the fiber network and the coax network in the customer premises. RFoG can be put in as standalone transmission medium or as an overlay system combined with EPON or GPON. In the last scenario WDM equipment multiplexes the downstream signals of RFoG and e.g. GPON on the fiber, which is possible as both transmission systems use different wavelength for up and downstream. In downstream direction services are broadcasted through the RFoG network just as they are in traditional HFC-networks. In upstream direction the CMTS controls the cable modem in exactly the same way as it does in an HFC-network. The cable modem, in its turn, deter- 16 Access Architecture Definition Document

17 mines when the RFoG transmits. Extremely responsive lasers react on the signal from the cable modem. The lasers are only on when the cable modem is transmitting. One of the advantages of RFoG is that it introduces very low noise in both upstream and downstream, this makes it possible to use higher modulation schemes enabling higher bit rates, or to bridge longer distances between the central office and the customer premises. Another advantage of RFoG over HFC networks is that fiber is less sensitive to humidity, temperature, lightning (EMC) than copper based networks. Compared to GPON or EPON systems, rolling out of RFoG technology is much easier and less expensive since a lot of equipment can be reused and operational processes can be unchanged, there will be no or very limited impact on the OSS and BSS systems. The drawback of the solution is that it is still based on the CMTS as generator/receiver for the digital information, as such the density and power / cost issues of the CMTS remain fully valid. With this solution the bandwidth of the fiber is not exploited as it could be, no differentiation can be made between a subscriber connected to the RFoG system compared to the HFC network. This draw back will certainly be important in case the operator has competition from an other FttH operator. Most likely RFoG or RF-PON will be used by a Cable operator at the beginning of the transition phase to a full fiber network. This way it offers the Cable operator at the start of implementing FttH to use the known systems like DOCSIS and DVB-C and the corresponding OSS and BSS. When the percentage of FttH customers grows, after a certain penetration degree the Cable operator can decide to introduce a full FttH solution including the corresponding OSS and BSS. With this scenario the investment costs per customer are lower. 4 Next Generation HFC networks Cable operators are basically standing before two kind of networks transitions to a NG-HFC network, see figure 6 Brownfield deeper Fiber Hub Hub Greenfield xpon figure 6: Topology of a Next Generation HFC network 1. Brownfield: In a brownfield case the cable operator has already an HFC network in place and the operator wants to have a transition in the most cost and technical effi- 17 Access Architecture Definition Document

18 cient way from his current network to a next generation HFC network, capable of offering more bandwidth. The cost modeling studies performed in deliverable D16 shows that the coax network is a strong added value and that it should be leveraged as much as possible, especially when digging costs are taken into account. 2. Greenfield: In a greenfield case there is not yet a network in place and the cable operator is completely free to design the NG-HFC network from scratch. In a greenfield situation the digging costs of fiber and coax should not be taken into account, because these costs have to be made anyhow. Fiber to the Home deployment is much more economical and future proof and therefore the correct way to go forward, as was concluded in deliverable D16. Given the fact that in a lot of European countries there is already a cable network in place with a good coverage, brownfield upgrades are important. Therefore the focus of this section is mainly on the brownfield situation, on how the operator can migrate to the NG-HFC network in the most cost and technology most efficient manner. The greenfield situation is discussed at the end of this chapter. 4.1 Upgrades from existing HFC techniques Deliverable D16 shows a summary of possible techniques that an operator can use in it s transition to the NG-HFC network. The first type of upgrades are derived nowadays from existing solutions but refined for NG usage: o o o Classical fiber node split, This architecture is based on dividing the Homes Passed behind an optical node over two or more mini nodes. All mini nodes have the same spectrum of the first node, but less homes passed have to share this capacity so more bandwidth is available per individual customer. The node split exists of implementing two or more nodes at the original node location. a node split does need more fiber capacity for connecting the optical nodes with the head end and vice versa. If there is no fiber capacity between these locations a WDM solution can be involved. This will increase the cost per homes passed. For future use the optical nodes should become dense to fit in the street cabinets and at the same time have less power consumption. Extension to 1GHz, Nowadays most HFC networks and all of its equipment and even the coax cables are designed for use till a maximum frequency of 862MHz. In theory the coax is not bound by the frequency of 862MHz but allows frequencies above. An extension to 1GHz can deliver extra capacity: 1000GHz 862MHz = 138MHz extra bandwidth. If this bandwidth is divided by 8MHz channels it supplies 17 extra channels, when using 256QAM it offers around 850Mb/s. But when using 4096QAM it can offer 1.4Gb/s. For future use this means that the HFC equipment should support frequencies till 1GHz. The biggest disadvantage of the Extension to 1GHz is that the networks and the equipment were not designed (in the past) to go above the 860MHz. Extension to 1GHz means that attenuation will be much higher in the upper spectrum than originally calculated, and thus an operator needs amplifiers with a usability to go till 1GHz and a higher gain. Full digital no FM, Full digital means that there will be no analogue signals (like PAL) in the complete spectrum of the cable operator. For future use this means that the HFC network can be redesigned for use with only digital signals, but also that the customer needs DA converters and/or STBs to use his legacy analogue equipment with the full digital network of the operator. 18 Access Architecture Definition Document

19 4.2 More disruptive scenario s Besides extending the already existing methods for increasing capacity in the network, the ReDeSign project also analyzed potential techniques to deliver interactive services. Again see deliverable D16 for detailed information. The main challenge for defining the NG-HFC equipment for interactive services is based on current costs and equipment, the operator will face huge cumulative investments costs and needs a lot of space in the CO or Hub for installing the equipment. For NG-HFC networks investment cost reduction and denser equipment is eminent to use the HFC network in the most efficient and valuable way. Based on current techniques, suitable to implement in a HFC network ReDeSign has done studies on the following possible NG-HFC solutions. o o o µ-cmts approach, The idea with the µ-cmts solution is to place small rugged CMTSs at the position of the optical nodes. This µ-cmts will serve one or more branches connected to the node. Because it is a CMTS it will contain the MAC functionality. The µ-cmts approach leaves the possibility to coexist with legacy services, but as the demand for capacity rises, eventually an operator can decide to go to full IP based on DOCSIS. µ-edge QAM approach, This solutions is based on the M-CMTS principle. It exists of one main Core CMTS based at a Head End or CO and Edge QAMs placed remotely at the position of the optical node. The Core CMTS will perform all the DOCSIS functionalities while the Tx- and Rx Edge QAM will perform the transmitting and receiving of the DOCSIS signals and the translation to Gigabit Ethernet. Like the µ-cmts the evolution of the network will start with a couple of Edge QAMs in the node, coexisting with legacy services and techniques. In the end there is no technical need to phase out digital and/or analogue services, but this solution offers the possibility to go to full digital. Because of the use of universal Edge QAMs in the node it is still possible to bypass the CMTS for digital TV services. Ethernet over Coax (EoC) solutions based on Access-MoCA / -G.hn / -Home PNA, This option is based on techniques that are used for IP connection within the home network. These techniques, if possible, can be extended for use in the access network of the operator. For the global description there is no difference between these three techniques, but the technique itself will determine if it s possible to extent the technique for use in the access network. Also the cost of certain techniques will influence the feasibility of using a certain technique. 4.3 NG-HFC MAC layer functionality In the previous paragraphs an overview has been given of different techniques, each with it s own (Medium Access Control) MAC protocol used to manage a shared medium. The reader is referred to deliverable D18 1 for detailed information and comparison of the different MAC technologies that are nowadays available to manage a shared medium infrastructure. In this chapter, based on the lessons learned from existing MACs a summary is given of what a MAC for a NG-HFC network should look like. Based on above information the technical requirements for a NG-HFC MAC solution will be specified on a basis of current techniques. The NG-HFC-MAC solution will not be specified in detail but a global description of the desired functionalities per technical aspect will be given: 1 In D18 the WiMAX MAC layer was also analysed. But because WiMAX is a wireless protocol, derived from DOCSIS and possible become obsolete, it will not be discussed in this deliverable. 19 Access Architecture Definition Document

20 4.3.1 Traffic Management Even in the NG-HFC MAC layer, traffic management will be the most important functionality of the MAC-layer. The amount of users in a shared segment of the HFC network may become smaller and smaller during the years, but the NG-HFC network will still be based on a shared medium based on broadcast technologies in the downstream and unicast technologies in the upstream. Because of the way the medium is build the NG-HFC network will always need a contention mode or period for new modems to notify themselves to the network. It will be a centralized master who will plan and grant access to the shard medium. Probably the allocation of bandwidth will be TDMA and FDMA based. S-CDMA still has to prove itself in the HFC network. The way to address the physical layer on FDMA can change due to use of channel bonding and/or different channel widths (e.g. as specified in DVB-C2). When the NG-HFC network doesn t have to support the legacy TV-channels anymore the channel width can become wider and thus decreasing the need for (complex) bandwidth techniques like channel bonding. When the total available amount of capacity is limited or has to be shared with many clients, the demand for Quality of Service (QoS) levels for different services will be more dominant than in networks were capacity is no issue at all. In these networks there will be less need to implement more advanced QoS levels, as long the service provider established the minimal requirements, like latency and jitter for the most critical services even on a Best Effort base. This will mean that the access to the medium and inherent to this the traffic management will remain an important functionality of the MAC layer. Especially when (future) applications or functionalities will become more critical with regards to bandwidth, timing, latency, jitter etc. Thus delivering the right Quality of Service will become necessary to optimally support these future services. Especially when the total amount of network capacity has to be taken into account, which is the case in a (NG-)HFC network that is upgraded following the demand curve. The NG-MAC layer should support at least 2 QoS classes, one QoS class with guarantees on transmission parameters and a Best Effort QoS class. The NG-MAC layer may support more QoS classes, but this should be balanced with the implementation complexity, manageability and usability of the QoS class by the cable operator. Looking to the described MAC protocols in this deliverable, it is clear that the MACs have each a different approach for the support of traffic management. Ethernet on the one hand has a very limited management of bandwidth and QoS, while on the other hand DOCSIS is fully optimized in terms of traffic management including an extensive support for QoS. Home oriented protocols like MoCA typically have lower support for QoS. Comparing PON MACs a strong difference is seen regarding bandwidth management and QoS, GPON is rather bandwidth efficient and has an extensive standardized QoS support while EPON supports 8 QoS classes and is less efficient Privacy and security Because of the shared character of the NG-HFC network, privacy and security will remain important. The NG-MAC layer should support data encryption based on AES (existing standard for encryption in access networks), support a secured way for exchanging encryption key information and perform checks on system integrity. Admission control is also a necessity in a modern access network. The MAC layer should take care of a security check in case new modems want to connect to the network. DOCSIS and MoCA support encryption based on DES which is considered nowadays to be insecure. GPON, G.hn and also DOCSIS support AES encryption which is the successor of DES. EPON does not have a standardized encryption, different vendors offer a propriety solution. Ethernet itself does not specify encryption at a MAC layer. All MACs that support AES encryption do support key management protocols for exchanging encryption keys and admission to the network. With MoCA the support for admission is ra- 20 Access Architecture Definition Document

21 ther weak and not sufficient for an access network. EPON and Ethernet do not standardize admission at the MAC level. DOCSIS support algorithm to perform checks on integrity of provisioning messages. GPON encrypt all control messages Digital efficiency Also in the future the bandwidth will stay scarce and valuable, see also paragraph 2. The need for digital efficiency will remain. The NG-HFC network should have an as small as possible needed MAC overhead and efficient allocations of the available bandwidth or time slots. The NG-HFC MAC layer should not waist any time slots on Back-Off time or idle time, so a tight as possible timing mechanism would be eminent. Piggy-back request and concatenation should be minimal supported. This implies that collisions should be avoided when delivering the services. It is not possible to completely eliminate the collisions. There should be a certain time interval necessary for modems to register themselves to the network. This period is when collision can and may happen. A solution with a centralized master, which controls the traffic seems to be the most logical solution for the NG-HFC network. Techniques to eliminate certain redundant information from higher levels should help to increase the Digital efficiency. Payload Header suppression, but maybe new techniques to transport more efficient streaming- or broadcast information should increase the efficient use of the available bandwidth. But the reader has to note that often the increase in digital efficiency comes at a cost of extra complexity of the MAC. Therefore the correct balance needs to be defined for every feature, how much is the efficiency improved versus how much complexity is added. DOCSIS and GPON are very efficient MACs. Ethernet is efficient in case the CSMA/CD is not used, otherwise efficiency under load is very bad. The efficiency of EPON is lower but acceptable. MoCA is due to its mesh structure rather inefficient. For G.hn this is not yet clear OSS services This is a difficult question for the NG-HFC MAC layer. Of course there should be a minimum functionality to support the appropriate (FCAPS) network management features. On the other hand more information from the systems makes it easier to deliver the necessary support to the customers for guaranteeing the Mean Time Between Failure (MTBF) of the services. A good example of multiple OSS features is the DOCSIS3.0 standard. This standard is very extensive and has besides standard SNMP MIB objects also the possibility to use IPDR (IP Detail Record). The big question will be if these additional, extensive features will be used in the future. Practically it will mean for a cable operator that an OSS software designer write applications to use this information. From the past there are examples of OSS feature embedded in equipment (CMTS) that are never been used because of lack of standardized OSS software to collect this information. Based on this experience it is clear that an OSS should fulfill a minimum functionality (e.g. FCAPS), but the added value of an extended feature set can be questioned. This remark is applicable for both existing HFC networks as NG- HFC networks. It should be noted that also for OSS services a layered approach is possible. MAC and equipment management should be supported by the OSS, as previously noted, but service management can happen in overlay, e.g. by using TR-69. All access MACs complies with the support of a MIB based OSS, more home oriented equipment like MoCA does not support this. For G.hn it is not yet clear. Ethernet has a number of additional standards specifying management. 21 Access Architecture Definition Document

22 4.3.5 General Considerations Of course the NG-HFC MAC layer will not have any unused features. Looking at the current technologies it will be clear that all services will be IP based, eliminating networks based on ATM and all support in the MAC for ATM. As explained earlier in this document complexity and the amount of mandatory functionalities will have a lot of influence to the general adoption of a standard by the industry. The Complexity of the NG-HFC MAC layer should be high enough to deliver the required functionalities required by the cable operators to deliver their services but also low enough to make it worthwhile for a vendor to take the specification and fabricate a product to offer to the cable operators, that is power and cost efficient. The NG-MAC layer should support at least the minimum functionalities, but is should be possible to extend the functionalities with optional standards. The vendor can extend the functionalities of his product based these optional standards. If possible, the NG-HFC MAC layer technology should be backward compatibility with legacy DOCSIS. This is not mandatory, but will make a transition from the current DOCSIS MAC to the NG-MAC easier to perform. This requirement should be compared to implementation complexity of the MAC. It might be more interesting from a cost and power consumption point of view to be not fully backward compliant, and to miss a number of use cases that are almost not used, but to bring down the power for all deployments. DOCSIS3.0 has a lot of unused features because of the legacy aspect of DOCSIS. DOC- SIS3.0 must be backward compatible with earlier versions. This results also that new introduces functionalities within DOCSIS3.0 will be used later when the equipment of earlier DOCSIS versions is phased out. The backward compatibility of DOCSIS is a great sales argument for current customers but makes the whole system complex. Other features that are mandatory in the DOCSIS specification like the various QoS configurations channel bonding, different modulation schemes etc. makes DOCSIS3.0 MAC layer the most complex technique of all. Often a standard specifies a very full set of functions ignoring the implementation complexity. A good example of this is GPON in comparison to EPON. GPON has a (too?) complete support of QoS and efficiency, while EPON has opted for a simple basic MAC with focus on the essentials. In reality operators can offer full triple play services with both technologies. Lesson learned from this is that it is key to define the correct feature set for a MAC without exhausting into features never used by an operator. GPON is inspired by DOCSIS, as a result of that the same concepts are applied but a number of unused features are not taken over. Ethernet over Coax protocols as MoCA and G.hn currently lack features in order to be usable in an access network. The simplicity of the Ethernet MAC, and the structure of the standard, makes it the least complex and most cost optimal MAC layer. The basic CSMA/CD concept can not be used in an access network due to the lack of efficiency. 4.4 Greenfield networks Greenfield, as explained earlier, means that there is no existing infrastructure that has to be taken into account. Nowadays the fiber techniques are mature and the prices of equipment and infrastructure are comparable with other legacy equipment and infrastructure. In all cases the digging costs are the same, therefore the operator will choose for the solution that offers the highest possible capacity at a cost effective price. For greenfield it will not be a choose between fiber and another technology but between different fiber topologies and associated technology. 22 Access Architecture Definition Document

23 5 The optimal NG-HFC network It is clear from the previous chapters that there is a large variation in the existing HFC networks out in the field, and that there are a number of possible technologies available or under investigation to upgrade the network. This chapter will bring all the information together, compare all possibilities, list the positive and negative points and make an advice on what a Next Generation HFC network could look like. In this deliverable the focus is on overview and summary, the reader is referred to other deliverables of the ReDeSign project for a more detailed analysis of each part. First the topologies for a NG-HFC network are discussed, then the technology options are compared, the cost aspect of the HFC network is separately discussed and finally a summary is made. 5.1 NG-HFC topologies All cable operators are evolving their network with architectures based on fiber closer to the user. The topology of the NG-HFC network will naturally bring fiber closer to the user. Most of today s HFC networks have still rather large fiber nodes, with sometimes relative long coax drop sessions to the user. A strong variation in network topology can be seen all over Europe. Due to the fact that the NG-HFC network should be an evolution from the existing HFC network, there will also be a variation in the topology of the future HFC network. The topologies of existing HFC networks are described in deliverable D06. Possible scenario s for the introduction of fiber into the HFC network, as a natural evolution of the network, is described in deliverable D09. In greenfield situations the most optimal NG-HFC topology will bring fiber to the user (FttU) or the building (FttB), because there is no network in place. In this case there is a clear advantage to choose optical fiber. The fiber is cheaper and more future proof compared to coax, on the long run. The possible topologies for this fiber network are discussed in deliverable D12. From a topology point of view there are three options, see figure 5: 1. a point-to-point architecture between every household and the hub (P2P architecture), 2. a point-to-point architecture between every household and an active element in the field (e.g. at the location of the fiber node) and a shared fiber between that point and the hub (active point-to-multipoint architecture between household and hub), 3. a point-to-multi point architecture with a full passive outside plant (passive point-tomulti point architecture between the household and the hub). The first architecture can only be used in small scale deployments, because the termination of all that fiber in the hub is not a scalable solution in case of large roll-outs. The second option solves this problem but the outside plant is still active, the last option is from a topology point of view the preferred one. It can scale to large deployments and the outside plant is passive, resulting in a better OPEX. Technologies to realize networks with this topology are discussed in the next paragraph. To summarize the most preferred outside plant in case of greenfield deployments is a pointto-multipoint passive fiber network. In brownfield situations fiber will be brought closer to the user in order to decrease the segment size (amount of homes passed) in order to increase the capacity of the network per user. Fiber has to be brought to the most economical point for that operator (Fiber to the most Economical point or FttE). Based on the cost analysis as reported in deliverable D16 it is clear that the digging cost of fiber is an important parameter in this discussion. Different possible scenarios of bringing fiber closer to the user have been described in deliverable D09. A first option is to split the existing fiber nodes without adding extra fiber. The segments 23 Access Architecture Definition Document

24 becomes smaller, increasing the net bandwidth per user, but fiber is not closer to the user. In most cases fiber will be brought to a point in the network at a location of an existing amplifier, such that the segment that is served by the coax network is small enough to be future proof in terms of bandwidth per user. Between this fiber node and the user still a limited number of coax amplifiers are being used. Some operators will drive this evolution to the extreme and go for a fiber to the last amplifier approach. In this case the fiber node is positioned in the network at the location of the last coax amplifier, the coax network between this fiber node and the user is passive. Each step in this chain described above has its advantages and disadvantages. In reality an operator will look at a business case taking into account the cost of the fiber roll-out, in most cases dominated by the digging cost, and the need for bandwidth in certain areas, determined by the actual network in place but also by the offering of competition in that area. Therefore this approach is called Fiber to the most Economical point. This point will be different for every operator; there is not one clear rule for all EU cable operators. To summarize in brownfield deployments there is not one single topology deployed by all cable operators for the realization of the NG-HFC network. But all variants of the topology will be based on point-to-multipoint HFC network with fiber closer to the user and with a last drop existing of a coax network that might be active or passive. 5.2 NG-HFC technologies This paragraph will discuss the technologies that are proposed or under discussion for the realization of the NG-HFC network, based on a topology as described in the previous paragraph. Different technologies have been analyzed within the framework of the ReDeSign project. Deliverable D12 discusses the fiber technologies that exist and that are possible candidates for the realization of the NG-HFC network. Deliverable D16 analyses a number of technologies that can be used in the coax part of the HFC network. Finally deliverable D18 makes an analysis of the MAC technologies for the organization of the point-to-multipoint architecture of the HFC network. For a detailed analysis or for background information on these topics the reader is referred to these deliverables. In this section the main results of the deliverables will be summarized, and guidelines will be formulated for the technology that is most suited for the realization of the NG-HFC network. From the previous paragraph it is clear that two topologies have to be analyzed. The first topology is a full fiber point-to-multipoint for the realization of a NG-HFC network in greenfield situations, the second topology is a mixed fiber-coax point-to-multipoint topology, with compared to the current HFC network a longer fiber span and smaller coax segments. In this paragraph first the fiber technology will be discussed, afterwards the coax technology will be summarized. Third the MAC layer analysis will be reported. Finally a number of full network proposals for the realization of a NG-HFC network in brownfield deployments will be analyzed and compared Fiber technology From the previous section it is clear that for greenfield deployments a point-to-multi point topology is preferable. Also in case of brownfield deployments such a topology has a number of advantages in case a deep fiber approach is followed. The optical point-to-multipoint topology is also called Passive Optical Network or PON, because of the passive nature of the network. Different technologies exist to enable this topology for the bidirectional transmission of data, as discussed earlier in this deliverable and in deliverable D12. One can distinguish mainly two approaches, the TDM-PON systems (with 2 standardized solutions EPON and GPON) and the RFoG PON systems. The reader is referred to deliverable D12 for a detailed analysis. The main difference between both approaches is in the MAC layer. For both approaches in downstream a regular multipoint data stream exist, as in an HFC network. In upstream a point-to-point connection between the end user and the CO or Hub is emulated by allocating timeslots to an ONU. The MAC layer is different. In case of RFoG the DOCSIS HFC MAC layer is reused over the PON network. The advantage is that this approach is 24 Access Architecture Definition Document

25 compatible with the existing HFC network, the drawback is that still DOCSIS equipment is needed for the MAC support and bandwidth available over the RFoG PON is similar as over an existing HFC network. TDM-PON systems are optimized for transmission of data over the PON network, as a result of that a more optimal MAC is available offering much higher peak bandwidths to users compared to RFoG, and the cost of the equipment is lower, as discussed in the next paragraph. The drawback is that this technology is new for a cable operator, often it does not fit with the existing OSS/BSS system. To summarize for small greenfield deployments RFoG is for a lot of operators the best way forward, the physical layer is ready for full fiber bandwidth on the long run, and on the short run there is no disruption with the existing infrastructure. For larger deployments TDM-PON systems are more optimal in terms of cost, power and density so are the optimal way forward for the cable operator. For the fiber part of the NG-HFC network, in case of very deep fiber architectures an RFoG or TDM-PON network offer advantages, in case of a regular extension of the fiber with deeper fiber nodes the current point-to-point approach can still be used Coax technology Not only the fiber part, but also the coax part in case of brownfield deployments has to be upgraded for a NG-HFC network. Deliverable D16 analyses a number of coax upgrade technologies. The reader is referred to that deliverable, and to a previous chapter of this deliverable for an in depth analysis. The main conclusions are recapitulated here. On the one hand a number of technologies have been analyzed for upgrading the existing HFC network, and increasing the capacity in the coax segment. Node splitting, extending the frequency range to 1GHz and analogue switch off or full digital are examples of that. At the other hand new coax transmission technologies have been studied and compared, like a µ- CMTS approach, a µ-edgeqam approach or Ethernet over Coax solutions. Finally a fiber to the last amplifier approach with the regular DOCSIS transmission has also been taken into account. A full comparison has been made in deliverable D06, the summary table is taken over in table 3. A number of conclusions can be drawn. First two techniques are looking interesting because of the high score on a lot of evaluation criteria, node splitting and FttLA. Node splitting is certainly a very good method to upgrade the HFC network, and that is why most cable operators are executing this. The drawback of the technique is that the capacity increase that can be obtained with node splitting is limited and that the problem of the density and power consumption of the equipment in the CO or Hub is not solved, on the contrary. As a conclusion node splitting will be mandatory in the upgrade of the HFC network, but on its own is not enough to transfer to the HFC network to a NG-HFC network. FttLA approach has communalities with the node splitting; it can be seen as an extreme result of node splitting. The bandwidth gain that can be obtained is of course very high, because the resulting coax segments are small. The investment that is needed is high, both for the role out of the fiber as for the active equipment that is needed in the CO or Hub to power all the segments with QAM channels. The power, cost and density problems of the CO or Hub are not solved. This approach has certainly a lot of potential and can be very attractive in case the digging cost of the fiber is moderate and the density problem in the CO/ Hub can be solved. Of course this technique can also be combined with e.g. a µ-cmts or µ-edgeqam approach. A NG-HFC network certainly will need a fiber part closer to the user, a FttLA approach is an extreme format of that so it certainly has to be looked at in the context of a migration towards a NG- HFC network. Extending the frequency spectrum and analogue switch off are techniques that in itself will not be enough for the creation of a NG-HFC network, but they might help reaching the required capacity in the network. Analogue switch off or full digital is on the roadmap of almost all operators, but on a very long timescale. Reality has to prove if acceleration of the analogue switch off is required or not. Extending the spectrum to 1GHz (or beyond) requires a strong investment in the outside plant, because most of the networks are designed for 860MHz. Increasing the spectrum would required the introduction of new amplifiers and the re-spacing of the amplifiers because of the higher losses above 860MHz. Therefore this 25 Access Architecture Definition Document

26 technique looks not attractive for most cable operators, based on balancing the investment that is required and the gain that is obtained. A new fact that has to be taken into account in this evaluation is the new amplifier technology has been invented in this research project. A new coax amplifier with 3dB more gain has been invented and demonstrated as part of the ReDeSign project. This technology, capable of supporting spectrum till or over the 1GHz, might make it possible to upgrade the coax network without the need for a re-spacing of the amplifiers. Of course the existing amplifiers need to be replaced by this new amplifier, but this upgrade is much easier, and thus cheaper, than a full re-engineering of the outside plant. To conclude this technique will not migrate the HFC network to a NG-HFC network on its own, but coupled with the new type of amplifiers it might be interesting as part of an upgrade scenario. Finally three other techniques have been evaluated, µ-cmts, µ-edgeqam and Ethernet over Coax. All three of them increases the capacity of the network, while at the same time tackle also the power, cost and density issues at the Hub. The drawback of the approaches is that they are more disruptive, they change part of the HFC network rather drastically. Node split extending to 1GHz full Digital micro CMTS Micro Edge QAM Access -Moca/ - G.hn / -HomePNA NG-FttLA Impact customer ++ +/ Impact network ++ +/ / coexistance with legacy services and techniques backward compatible with legacy services and techniques Capex Investment ++ + // /- +/ capacity gain Downstream +/- -- +/- +/- +/ capacity gain Upstream +/- -- +/- +/- +/ availability easy migration / Equipment density CO table 3: Comparison of different upgrade technologies Different technologies have been compared with each other; most of the technologies do not stand on its own but can be combined with each other. Remains the question, what is the optimal way forward for the realization of the NG-HFC network, from a coax transmission point of view. At this moment it is not possible to bring forward one technology that realizes the NG-HFC network that is optimal in terms of cost, power and density, that is not disruptive and that has a very easy migration path starting from the current HFC network. There are mainly three scenario s which are illustrated in figure The first solution is called Dense Hub (or CO). This solution is closest to the existing HFC network. The network is based on a fiber closer to the user topology (in extreme FttLA). Transmission of information over the network is the same as in the existing HFC networks, analogue TV, digital TV and data via DOCSIS. The bandwidth is increased by bringing fiber deeper in the network and by the creation of smaller segments. There is a lot of DOCSIS equipment required in the Hub in order to support all these segments. Based on the current CMTS and QAM equipment this will give a high cost and power issues together with density problems. The solution for these problems put forward is a next generation of CMTS equipment that is much denser in terms of QAM channels compared to the existing equipment practice. The fixed ratio between US and DS QAM channels will also be removed. This solution is currently under investigation by Comcast in the USA which is the main driver for this architecture. The new generation of CMTS equipment does not exist at this moment, vendors are investigating possibilities. 26 Access Architecture Definition Document

27 figure 7: Three most probable architectures for the realization of the NG-HFC network. 2. The second proposal is called µ approach. This solution is based on pushing functionality out of the Hub, into the field. One can distinguish two variants. The first approach is an extension of the modular CMTS architecture. The QAM modulators are pushed in the outside plant, at the location of the fiber nodes (positioned deeper in the network compared to today in order to generate smaller segments). The M-CMTS is still in the Hub. The second variant pushes the full CMTS functionality into the field, a µ-cmts is defined containing the functionality of an I-CMTS, but scaled down to the dimensions of one (or a few) segments. Both approaches are logical evolutions of the current HFC networks. The architecture as defined by Cable Labs of the M-CMTS is extended. The density and power issues are solved by pushing functionality out of the Hub into the field, to a position (fiber node) that already needs power. The problem with the first approach is the timing between the M-CMTS and the µ-edgeqam. It is unknown if the solutions of today are robust enough in terms of timing to support a long distance fiber network between both devices. Second problem is the US, that is not solved in a nice way with the M-CMTS EdgeQAM architecture. Therefore a µ- CMTS approach, a device that does not exist yet, will solve both issues, and also the density and power problems of the Hub. The µ-cmts is a downsized version of the regular CMTS, higher volumes and lower dimensions should bring cost down. 3. The third solution proposed can be summarized under the name Ethernet over Coax. This proposal is not a single solution, but is a group of technologies that have in common that they propose a different transmission mechanism for data over coax. The dominant protocol for data transmission is Ethernet, therefore these solutions are called Ethernet over Coax. Within this group different proposals exist for the transmission of the information over the coax network. Most of the proposals are inspired by transmission technologies used or over different media (e.g. Wi-Fi, ultrawideband) or in different parts of the network (e.g. the home). The technologies are ported to the coax access network and are adapted to that specific network segment. A number of vendors have propriety solutions available on the market, a number of them are currently in field trial, mainly in Asia. The strong part of these proposals is often that 27 Access Architecture Definition Document

28 components for PHY transmission are reused from other network segments, increasing the volume and lowering the cost of the solution. The density and power issues in the Hub are also really solved, a classical edge routed in the Hub can do the job. The weak part of the proposals is that there are a number of proposals out, each propriety from a single vendor, without standardization. The solutions are also not yet field proven for mass deployments. To conclude on the coax technology, mainly three options are available. None of the options are available on the market for the cable operator yet. The options differ strongly in how close they are aligned to the existing HFC network. Operators that really want to be as close as possible to the existing HFC approach might be interested in proposal 1. Operators that still prefer an easy migration path without too much disruption, but are really concerned about the density and power issues in the Hub might go for solution 2. Operators that are less concerned about the backward compatibility of the NG-HFC network, and that are open for the introduction of new technology might choose for proposal 3. For the European cable operators the second approach might look the most attractive solution, with the introduction of a µ-cmts in the network. The introduction of the µ-cmts could also support DVB-C2 as PHY layer, as described in other deliverables of this project. As such a powerful, future proof and scalable solution for the realization of the NG-HFC network becomes available. North American cable operators are often less concerned about power and density, but are very concerned about the introduction of new technologies, so proposal 1 will probably match there needs. Asian cable operators are less advanced at this moment, they do not have yet made such a big investment in DOCSIS equipment as compared to American or European cable operators. Probably they have a clear look at proposal MAC technology All three proposals as described in the previous paragraph are point-to-multipoint networks, two of the three proposals are still based on DOCSIS or DOCSIS evolutions. Therefore the MAC functionality, enabling the point-to-multi point PHY layer, is rather important for the creation of en efficient NG-HFC network. Within the ReDeSign project the MAC functionality has been described and different MACs (DOCSIS but also others) have been analyzed and compared, see also paragraph 4.3. The work has been reported in deliverable D18. The main conclusions and results will be recapitulated in this paragraph. Comparing different MACs as described in this deliverable one can see a strong difference in terms of functionality and philosophy. A number of MACs are focused on the basic functionality while other MACs are specifying a lot of functionality. It is not the intent of this deliverable to label a MAC as good or bad. For the definition of a possible NG-HFC MAC the features of which we think are important are listed below, taking into account on the one hand the functional behavior of the MAC, but on the other hand also the impact on the implementation. An overview of the MAC analysis is given in table Access Architecture Definition Document

29 DOCSIS MoCA G.hn GPON EPON Ethernet NG-MAC Traffic Management Privacy and security Digital Efficiency ++ -? ++ +/ OSS Services ++ --? Complexity / unused features / table 4: Comparison of the different MAC layers with the NG-HFC MAC. Taking into account all considerations summarized in chapter 4.3 a number of conclusions can be drawn towards a NG-HFC MAC. It is clear that a cable operator will not introduce a completely new concept in his network, so the NG-HFC MAC should start from the current DOCSIS MAC. DOCSIS itself has a lot of useable functionalities and is best suited to offer support to the Cable operator to deliver the required services. On the other hand DOCSIS is a very complex MAC, certainly compared to other MACs that also enable triple play services over a shared medium. An effort should be made to reduce the complexity in areas that are less critical for the cable operator. Moreover one has to think of introducing the Ethernet concept, define a basic mandatory standard extended with optional functions. This approach will also open the market for more vendors to enter the market, resulting in lower prices. To conclude two out of thee proposals for the creation of the coax transmission of the NG- HFC network, can be best supported by an evolution of the DOCSIS MAC. The DOCSIS MAC should evolve as described above, a number of features should be left out and others should be made optional. A reduced DOCSIS MAC in terms of functionality will strongly helps to keep the cost of the µ-cmts concept low. The third proposal (Ethernet over Coax) for the creation of the NG-HFC network also requires a MAC layer, but backward compatibility with DOCSIS is less a concern. A new MAC concept might be introduced, but it should fulfill the application and service requirements of the cable operator. Therefore it might be a good idea to also for this proposal to adopt an evolution of the DOCSIS MAC. 5.3 Cost evaluation Up till now the NG-HFC network has been discussed from a technical point of view. Another important parameters is the cost. Within the ReDeSign project the cost of the HFC network based on the technology of today is compared to a FttH network based on PON technology of today. The approach taken for the cost evaluation and the results are discussed in deliverable D Cost comparison of HFC and FttH For this cost analysis it is key to understand the simple model that has been used, and the reason why it has been used. Cost modeling and looking toward the future is always a difficult business with inherent rather high error margins. Therefore it is important to understand and interpret the results carefully. The reason of the error margin is twofold, on the one hand there is the error in the model. Depending how accurate you make your model you can have impact on this. Second are the input values used for the calculations, which are in this case the costs taken into account in the model. The total error margin is the combination of both. The last aspect that has to be taken into account the aim of the cost model, what do you want to obtain with the results and how sensitive are the results and the interpretations on changes in the input parameters or in variations in the results. 29 Access Architecture Definition Document

30 Let us have a closer look at the cost model as discussed in deliverable D16, and the considerations above. The simple model has the drawback that there is a large error margin on the result, and that the results have to be carefully interpreted. Another approach is to make a rather realistic model. In this case the network model will be made much more complex, taking into account a lot more of the parameters that in reality will have an impact, such as e.g. the evolution of the equipment, upgrades in the outside plant, etc. The problem with this approach is that it is very difficult to obtain consistent, correct input values for this more complex model. The result of the model may seem more accurate, but in reality the error margins on the input translates into high error margins on the output. Therefore in this project the choice has been made to go for a simple cost model, but with rather correct input values. As such the error flag is dominated by the error flag of the model, which can be understood and correctly interpreted. Also for the reader it is clear that the results should be interpreted and not taken over as such. For our case it was possible to find input values for all parameters of the model (except the digging cost) in one source (a study of Dell Oro). The actual values of the results are not important for the interpretation, but the relative values of the HFC network compared to the FttH network are important. This reduces the impact of the error flag from the input parameters on the actual results. Out of the cost calculations a number of conclusions can be drawn. First the results are given for the normal case, and the interpretation of the results is taken over from deliverable D16. The cost of the HFC network is compared with the cost of the GPON based network. The cost figures are calculated based on the cost model as discussed above. Costs have been calculated for a period of 10 years. In a case illustrated here the total investment cost in a certain year has been calculated, as a function of time for 4 cases: a GPON network with 128 users per PON, a GPON network with 32 users per PON, an HFC network with a BW growth of 60% per year and an HFC network with a BW growth of 40% per year. First results are shown in figure 8. Out of these results a number of conclusions can be drawn. The investment for upgrading an HFC network today is much smaller than for deploying a GPON network. The total cumulative investment in an HFC network grows gradually, with the increase in bandwidth. The investment in a GPON network will also follow the bandwidth, but with a step function. A relative high investment cost is required in order to connect all users. This network has a lot of excess bandwidth compared to the demand. From the point that the GPON network would not offer enough bandwidth anymore, a new investment will be necessary, resulting in an increase in investment for a short period, followed by again a period without a lot of investment required. Over a long time span the investment curve for GPON will be a step function, each time the capacity of the PON is not sufficient anymore a step in the curve will be seen, followed by a flat curve for some time. This would e.g. be the case for the deployment of a 1:128 GPON system, in 2018, in case the bandwidth demand is increasing every year with 60%. The investment in a 1:32 split or a 1:128 split GPON network does not differ so strongly. The small increase every year for the GPON cases is due to the increase in subscribers every year. New equipment, like ONTs, will be required to connect these new subscribers. Over time (for our assumptions after 2018) it will be cheaper to build a full new FttH GPON network, including digging of the fiber, compared to upgrading the HFC network. 30 Access Architecture Definition Document

31 Investment [$] 40% growth 60% growth GPON 1:128 GPON 1: figure 8: Total investment versus time for an HFC network and a GPON FttH network, with average cost for digging included The total investment taken into account for these cost calculations is based on the active equipment in the HUB, at the customer premises and for GPON the digging of the fiber. Special parameter in this comparison is the cost of the digging. There are two aspects on the digging cost, first the actual cost varies strongly from country to country and second sometimes this cost should not be taken into account because it is not part of the business case (e.g. for greenfield deployments, or in case digging cost is subsidized by a third party (Government or others)). The results as given in figure 8 are created with an average cost for digging included. The case without digging cost is also interesting as reference and is illustrated in figure 9. Investment [$] 40% growth 60% growth GPON 1:128 GPON 1: figure 9: Total investment versus time for an HFC network and a GPON FttH network, digging cost not included From the comparison between figure 8 and figure 9 some conclusions can be formulated: 31 Access Architecture Definition Document

32 The investment required for a GPON FttH network is strongly determined by the digging cost of the fiber. In case digging cost is not part of the business case (e.g. greenfield area) FttH GPON is within a few years time economically more attractive than HFC. In case of a high bandwidth growth (60%) per year, GPON will be far cheaper than HFC in 10 years from now. figure 10: Topologies for a FttLA approach: high density Cost of FttLA One last aspect that has not been discussed in deliverable D16 is the cost of a FttLA approach. This network topology is rather attractive for a lot of cable operators because the CAPEX investment that has to be made is lower compared to full FttH, for the customers there is no change (in case DOCSIS is still used) as no replacement of the modem is required. Also the OPEX cost of the operator is going down because the network between the node and the customer premises is passive, no amplifiers anymore. The main investment that has to be made is of course the creation of the fiber network, dominated by the digging cost. This cost is strongly depending on local conditions, and is therefore not taken into account here. Other cost for the creation of this network, are the costs of the optical nodes. Much more optical nodes will be required compared to a classical HFC network. Two cases have been analyzed, a first case is a dense urban network illustrated in figure 10, a second case is a medium density case see figure 11. For each topology two options are discussed, the first option is a full FttLA approach with no amplifier between the optical node and the modem. The second approach considered in this analysis contains still 1 amplifier between the optical node and the modem located at the customer premises. For this architecture an amplifier of the type place and forget could be used. 32 Access Architecture Definition Document

33 figure 11: Topologies for a FttLA approach: low to medium density The results of this analysis are given in table 5. Two types of nodes have been defined, a node with 1 output (serving one coax segment) and a node with 2 outputs (serving two coax segments). In total eight cases have been analyzed. First a distinction has been made in the density of the network. In case of a high density network more users are connected to a segment compared to a low to medium density network. Second distinction that has been made is in the topology of the optical node. Finally a true FttLA approach has been analyzed, with a passive coax network, or a semi FttLA approach with still one coax amplifier used in the coax segment. The cases that have been calculated are illustrated above, and are based on realistic deployments in streets. In figure 12 the different topologies used for the cost analysis are illustrated. Please note that the numbers in table 5 for homes passed per optical node are average values based on the realistic segmentation of the street deployments as illustrated above. The cost that has been calculated in the model is the cost per homes passed. The cost of the lowest case has been labeled 100% and the cost of the other cases have been given relative to the lowest case. The first conclusion that can be drawn is that a true FttLA approach is more expensive compared to the case with still one coax amplifier in the coax segment. The cost difference of an optical node supporting one segment or two segments is not that big. 33 Access Architecture Definition Document

34 figure 12: Different FttLA topologies as used for the cost calculations. ARCHITECTURE LOW/MEDIUM DENSITY Node 2 active outputs Homes Passes / Optical Node Mb/s / Homes Passed price / Homes Passed 1 active output 1 active output + 1 amplifier 2 active outputs + 1 amplifier HIGH DENSITY 2 active outputs 1 active output 1 active output + 1 amplifier 2 active outputs + 1 amplifier ,21 99,85 33,28 24,96 52,35 71,98 22,90 16, % 275 % 111 % 100 % 249 % 294 % 127 % 100 % table 5: Results of a cost analysis of the FttLA approach 5.4 Proposals for NG-HFC architectures This section will summarize the conclusions of the ReDeSign project concerning Next Generation HFC networks. As explained in this chapter a distinction has to be made between brownfield and greenfield deployments. Both cases are discussed Brownfield In case of a brownfield deployment, a cable operator will start from a current situation, there is already an existing HFC network in place. In this case the cable operator is advised to upgrade his HFC network towards a next generation HFC network. The timeline of this upgrade will depend on local factors such as service uptake, bandwidth demand and competition. Upgrading the HFC network will include brining fiber closer to the user. The actual optimal 34 Access Architecture Definition Document

35 point to where fiber has to be terminated is determined by the digging cost of the fiber (FttE or Fiber to the most Economical point), which has to be carefully analyzed by the operator for his network or network segment. In the exceptional case that there is no digging cost to be taken into account (because fiber is already available or the cost is paid by a third party), the upgrade has to be looked at as a greenfield approach. In case of low digging cost a FttLA approach can be attractive. In case of a higher digging cost the optical node will be positioned deeper in the network. It can be interesting to upgrade the current point-to-point fiber network between the hub and the optical nodes to a PON network in case a lot of optical nodes have to be deployed (e.g. FttLA topology). The coax part of the network needs also an upgrade. For European cable operators probably the most suitable network approach is an evolution from DOCSIS, and the introduction of a new type of optical node integrating also the PHY and MAC layer of the coax transmission, the µ-cmts approach. This will solve the density issues in the hub. In order to optimally accommodate for this network approach, a new version of DOCSIS is preferred, including new evolutions in the PHY layer to increase the available bandwidth (DVB-C2) and evolutions in the MAC layer, to reduce the complexity and cost of the MAC implementations Greenfield In case of a greenfield approach, there is not yet a network in place. Although coax offers a high bandwidth that certainly will be capable of supporting the bandwidth demands of the customers for quite some time, due to CAPEX and OPEX cost aspects it will be preferable to introduce fiber all the way to the home (FttH) or building (FttB), digging cost will be the same for coax or fiber and the cost of the fiber is lower than that of coax. In case no digging is required but existing tubes can be used, blowing fiber is cheaper than coax, so also here there is an advantage to go immediately to fiber. So from a physical transport medium point of view fiber is much more attractive in this case. From a MAC layer point of view (how will I use the fiber) there are mainly two options. The advice is that in case the greenfield deployment is small and is in the neighborhood of existing HFC network of the cable operator, the use of the RFoG technology is preferable. With this approach the MAC-layer of DOCSIS is reused over the fiber. In case the greenfield deployment is larger or in case there is not yet an HFC network in the neighborhood a TDM-PON (GPON or EPON) approach is preferred. In this case a new MAC, but also new equipment with its OSS/BSS system is introduced in the network, but on the longer run CAPEX and OPEX savings will be high enough to make this approach by far more attractive. 6 Conclusions A number of conclusions can be drawn from the work carried out in the ReDeSign WP6 project, and reported in this deliverable. First conclusion is that on the longer run there is a real need for a NG-HFC network by the cable operators, normal upgrading techniques will not provide the required bandwidth growth and associated service support in the most efficient way. Coax as transmission medium has still a long way to go, the capacity of the transmission medium is very large and future proof. The point is that the transmission of the information over the network should be scalable and cost/power efficient with high density equipment. This brings the NG-HFC network in the picture. One has to make a distinction between brownfield and greenfield deployments. In case of greenfield deployments the most optimal way forwards is the roll-out of a full fiber network till the building or home (FttH or FttB). Roll-out of coax does not make economical sense. Technology to be used to enable the fiber network will depend on some local specific cable operator parameters. A choice has to be made between TDM based PON systems (EPON or GPON) or RFoG technology. 35 Access Architecture Definition Document

36 In case of brownfield deployments the definition of the optimal NG-HFC network is less obvious, but all scenario s will reuse the current coax outside plant as maximal as possible. Study shows that the main asset of the cable operators is the coax outside plant, investments in a new network (e.g. fiber) are very high. Depending on the installed base (e.g. investment made in DOCSIS equipment, actual size of the Hubs, CO, HE) one or another solution might be preferred for the realization of the NG- HFC network. Mainly three options can be put forward. The first option is a natural evolution of the current HFC approach, based on fiber closer to the user to reduce the segments and combined with a new kind of CMTS equipment that is much more optimized and dense compared to the current equipment practice. This solution is called the dense Hub or CO option, with some traction in the USA. The second option is a solution based on fiber closer to the user combined with pushing the DOCSIS functionality also closer to the user. This solution will give optimal advantage in case a new version of DOCSIS can be standardized tuned to this deployment scenario. This solution is called µ-cmts or µ-edgeqam and might be of interest for European cable operators that already made a lot of investment in DOCSIS equipment. The third and last option introduces a new way of data transmission over coax (not using DOCSIS anymore) called Ethernet over Coax and combines this with a deep fiber architecture. This solution is called EoC and is trialed by a number of Chinese cable operators. Considering these options in the light of cost (both CAPEX and OPEX), power consumption, technique, services, QoS and equipment density the second option might look most optimal for European cable operators. This option will rely a lot on an evolution in DOCSIS standardization, in order to optimize the proposal for lowest cost at maximum performance. A new version of the DOCSIS MAC, a lightweight version of the current standard, will improve the implementation complexity, and as such the associated cost. In case the evolution of DOCSIS will follow the existing track, the density issues in the CO and investment costs for DOCSIS equipment will become a reason for cable operators to look at alternatives, like Ethernet over Coax or will make the business case for a cable operator to move to FttH positive in an earlier timeframe. 36 Access Architecture Definition Document

37 Annex 1: List of abbreviations AE Active Ethernet AES Advanced Encryption Standard CAPEX Capital Expenditures CDMA Collision Detect Multiple Access CMs Cable Modems CMTS Cable Modem Termination System CSMA Collision Sense Multiple Access CPE Customer Premises Equipment CO Central Office CWDM Coarse Wavelength Division Multiplexing DES Data Encryption Standard DOCSIS Data Over Coax Service Interface Specification DWDM Dense Wavelength Division Multiplexing DVB-C Digital Video Broadcast Cable EoC Ethernet over Coax EPON Ethernet Passive Optical Network ertps Extended Real-Time Polling Service FCAPS Fault, Configuration, Accounting, Performance, Security FDMA Frequency Division Multiple Access FM Frequency Multiplexing FSAN Full Service Access Network FttH Fiber to the Home FttB Fiber to the Building FttE Fiber to the (most) Economical point FttLA Fiber to the Last Amplifier GPON Gigabit Passive Optical Network HE Head End HFC Hybrid Fiber Coax HP Homes Passed HPNA Home Phoneline Networking Alliance IP Internet Protocol LC Local Centre MAC Media Access Control MIB Management Information Base MoCA Multimedia over Coax Alliance 37 Access Architecture Definition Document

38 MPCP MPEG MSO MTBF NG-HFC OFDM OLT ONU ONT OPEX OSS P2P P2MP PAL PHY PON QoS QAM RF RFoG S-CDMA SECAM SNMP TDM TDMA WDM Multipoint Control Protocol Moving Pictures Expert Group Multiple Service Operator Mean Time Between Failure Next Generation HFC Orthogonal Frequency Division Multiplexing Optical Line Termination Optical Network Unit Optical Network Terminal Operational Expenditures Operational System Support Point-to-Point Point to Multi-Point Phase Alternating Line Physical layer Passive Optical Network Quality of Service Quadrature Amplitude Modulation Radio Frequency RF over Glass Synchronous-Code Division Multiple Access Séquentiel Couleur à Mémoire Simple Network Management Protocol Time Division Multiplexing Time Division Multiple Access Wavelength Division Multiplexing 38 Access Architecture Definition Document

39 Annex 2: References [1] Deliverable D03, HFC Architectures Questionnaire Report, version 1.0, June 2008 [2] Deliverable D04, Service requirements report, version 1.0, June 2008 [3] Deliverable D07, Technical requirements report, version 1.0, December 2008 [4] Deliverable D09, Topologies short list report, version 1.0, February 2009 [5] Deliverable D06, Reference architectures report, version 1.0, October 2008 [6] Deliverable D12, Fibre transmission technologies short-list report, version 1.0, June 2009 [7] Deliverable D16, Topology and Transmission Technology Details Document, version 1.0, November 2009 [8] Deliverable D18, MAC layer short-list, version 1.0 April 2010 [9] 39 Access Architecture Definition Document