The M-CMTS architecture could be the transport answer to burgeoning customer data capacity and throughput demands. As Internet protocol (IP) continues to proliferate at an increasing pace throughout cable operators’ networks, it is clearly becoming the standard end-to-end transport mechanism to enable true network convergence at all layers. And with lines blurring among telecommunications providers, Web content aggregators and cable operators, service integration and the associated cost reductions are becoming more critical than ever to cable operators’ success. Industry specifications and architectural guidelines lay out a roadmap to IP-enabled convergence, but recently the emergence of interim proposals has left some wondering exactly when these new architectures will come to fruition. Background Now more than ever, consumer demand is driving innovation and market entry timelines. Cable operators will not be quick to change simply for the sake of changing—nor should they be. Customers today don’t care whether “The Sopranos” is piped in over IP or MPEG, but consumers will demand and adopt change when providers can leverage the power of IP convergence to create truly unique services not possible with MPEG or other broadcast technologies. Any change at the network level has to add value for the customer, while also adding value to the service portfolio and reducing operational costs. Many interesting technologies today cannot readily be implemented because of the high cost of integrating with the existing proprietary infrastructures, as well as lengthy development and implementation challenges resulting from non-standards-based designs. Today, cable operators are focusing their convergence efforts on the network layer, breaking down the legacy silos of different transport technologies in an effort to create open, standards-based networks that can support a multitude of services much more efficiently than a single network supporting each individual service type. In the end, there are several layers to complete convergence, but the initial phases at the network level need to be completed first in an effort to achieve the desired “single network-multi-service architecture.” The edge network While cable’s competitors are busy building completely new networks from the ground up to provide triple-play services, cable operators have the advantage of taking a more deliberate approach by leveraging existing networks while migrating to all-digital and all-IP environments. One of the huge benefits of the historical investment in the HFC infrastructure is in its transparency. Because the HFC network provides tremendous flexibility and extensibility, operators can simply modify the end points to meet changing service requirements, as opposed to continuing to build out the last mile network. Platforms moving forward must mirror this flexibility to improve service velocity to accommodate the rapidly changing demands of the customer. Operators today cannot afford equipment with limited shelf lives that require replacement because of lack of extensibility. Developing modular edge network elements that can flexibly scale and adapt to changing network requirements appears to be the preferred strategy to ensure cable operators can continue to leverage the extensibility of the HFC infrastructure to their advantage. The modular cable modem termination system (M-CMTS) is a concrete example of equipment vendors and cable operators coming together to define the requirements for a platform that will support these goals well into the future. By decoupling the physical layer (PHY) from the existing integrated CMTS, cable operators will be able to leverage single quadrature amplitude modulation (QAM) technology/platforms for video, voice, and data services. Technology leaders are extending modular philosophy to the network routing edge, which, because of the complex switching and routing requirements in today’s networks, has historically stood alone. However, Moore’s Law continues to provide advances in silicon (albeit an expedited version), and technological breakthroughs will allow these complex algorithms to be managed while consolidating other processing needs such as DOCSIS, video switching, deep packet inspection (DPI), and others into one platform. This creates a benefit to the operator that provides efficiencies and synergies both operationally and economically. First steps As they begin their transition to modular edge devices, cable operators will seek to leverage existing platforms as much as possible to minimize operational burdens and expenses. The plans to test and deploy pre-DOCSIS 3.0 channel bonding solutions showed the desire of some operators to bring a subset of the DOCSIS 3.0 features to the market sooner, such as channel bonding, advanced multicast capabilities, and IPv6. In the near term, operators will focus on achieving the documented M-CMTS goals: *”Deploy independently scalable numbers of downstream channels without changing the MAC (media access control) domain or the number of upstream DOCSIS channels”; and *”Lower the cost to deliver video over DOCSIS service to be competitive with today’s MPEG VOD (video on demand) by implementing a new generation of `de-coupled’ downstream-only cards on existing CMTSs.” With “coupled” downstream/upstream CMTS line cards, the incremental cost for a DOCSIS downstream channel is excessive because of the need to purchase unnecessary upstream ports. For example, assuming a two-down/eight-up line card costs $20,000 means that the incremental cost of increasing only downstream capacity is $10,000 per QAM channel. Contrast this with the incremental cost of an MPEG-2 downstream channel, which, at the time of this writing, was approximately $500 per QAM channel, and the cost savings of de-coupling the up- and downstream cards are attractive. It is expected that introducing a “decoupled” downstream-only card into an existing CMTS chassis can bring the incremental cost per QAM channel of the decoupled card much closer to the cost of an MPEG-2 channel for a full DOCSIS downstream channel that supports baseline privacy initiative (BPI) encryption, payload header suppression, and DOCSIS per-service-flow quality of service (QoS) scheduling. Thus, the primary goals of M-CMTS—decoupling and lower cost—are met merely by introducing a decoupled downstream CMTS blade. Next steps: IPTV? In the future, when a significant portion of DOCSIS downstream throughput becomes IPTV and data rate requirements grow beyond what can be met by an integrated CMTS, an M-CMTS architecture could provide the flexibility to meet growing throughput demands. Cable operators may reasonably wonder: “Why bother with IPTV?” After all, MPEG-based digital video for broadcast and VOD is cheap, reliable and ubiquitously deployed. As depicted in Figure 1, IPTV supplements their existing MPEG video services by increasing the number of video sources and similarly increasing the number of video playback devices. Millions of new Internet-based video titles are becoming available to subscribers via broadband IP connections. These include streaming Webcasts and download to view/own movies. In addition, there is a proliferation of display devices (or home terminals) beyond the standard set-top box, such as game consoles, video phones, and PC video displays. Many of these are wireless, and they all operate over the home IP network. What is the likely required “endgame” data capacity and throughput for IPTV in five to seven years? Consider a scenario in which any digital set-top box or high-speed data subscriber can receive VOD of “What I Want When I Want” (WIWWIW). Let’s use the following assumptions: * 750 homes passed per fiber node and a penetration of 75 percent for either set-top box or cable modem * 50 percent of subscribers are concurrently watching WIWWIW video, meaning that the content is not pre-stored on the subscriber’s DVR * Two standard-definition (SD) and one high definition (HD) MPEG4 streams per home, equaling 10 Mbps per home The total throughput per fiber node is 750 households x 75 percent penetration x 50 percent concurrent usage x 10 Mbps bandwidth = 2,813 Mbps, or 73 256-QAM downstream channels’ worth of unique content. This is an upper bound on the number of unique 6 MHz channels required for VOD with 750 HHP fiber nodes. Splitting the fiber node to fewer households passed will reduce the number of unique channels, but does not reduce the overall capacity required for all 750 homes, and this example only considers one node within a headend. Contrast that example with an “endgame” scenario for high-speed data that calls for 50 percent penetration of DOCSIS cable modems with a 100 Mbps rate cap (i.e., all modems are DOCSIS 3.0 or later) and an engineered peak-to-average of 0.25 percent concurrent usage. The total data rate for a 750 household passed node is 750 households x 50 percent penetration x 0.0025 concurrency x 100 Mbps bandwidth = 94 Mbps. Thus, the endgame throughput for VOD is expected to be more than 26 times the data rate for high-speed data. Given the popularity of Internet IP video sources and IP video displays, as well as deployment of IP set-top boxes, much VOD traffic could be transmitted as IPTV over DOCSIS, rather than as digital video MPEG over nonDOCSIS QAM channels. Though some emerging arguments consider using DOCSIS 100 percent of the time for IP video distribution to the home (potentially burdening IP video streams with BPI gates designed for data services), the likely scenario will be a combination of DOCSIS and a pure IP transport method for video. This then creates a true end-to-end IP distribution method that is flexible in design, setting up the ability to increase service velocity. Completing the migration As an industry, separating the MAC and PHY portions of the existing network is a huge step toward the goal of achieving a robust, open, and scalable architecture; however, it is really only the beginning. Another challenge is the need for robust L2/L3 aggregation as part of this next generation access design. The metro networks of today are comparable to the core networks of 6-7 years ago. Consider, for example, that many metro networks are now offering multiple 10 Gigabit Ethernet (GigE) links tied together in ring or even point-to-point architectures to support increases in HD content, the rapid adoption of VOD, higher speed data tiers, and eventually an all digital environment. Today’s duplication model of adding a CMTS when a capacity threshold is realized can be inefficient from a cost, operational and facilities standpoint and is giving way to a more consolidated solution. These new solutions will require best-of-breed edge routing platforms to support quality, availability and granular control of future services. It is advantageous for operators to utilize existing platforms in an effort to reduce stranded assets in the local systems. However, there may be a natural breakpoint that eases the transition to a fully integrated M-CMTS architecture. While early applications may be few that have the ability to utilize the 100 Mbps+ downstream capability that DOCSIS 3.0 will offer in the near-term, with expanded video services and other data and voice applications emerging, the time will come when local markets may reach a capacity threshold with existing integrated devices that are running such features as channel bonding. At this demarcation point, the historical paradigm of simply adding an additional stand-alone CMTS will give way to a migration to a true M-CMTS architecture. This sense of urgency was made evident in the short-lived DOCSIS 2.0b proposal. Although recently voted down by CableLabs and its members to ensure a continuous and concentrated focus on the development of DOCSIS 3.0 product, the fact that the issue was raised confirms the desire and eventual need for a transition to this new and robust architecture. Figure 2 is a graphical representation of this design, using interfaces ratified by CableLabs. In Figure 2, the original idea of a stand-alone MAC core has given way to a more consolidated model, combining a new breed of edge routing platform that is IP, DOCSIS and application aware, made possible in part by the M-CMTS specifications. Combined with existing services, service convergence continues to drive the need for network convergence, creating a requirement for platforms to support existing and emerging services such as digital simulcast, DOCSIS Set-Top Gateway (DSG), switched digital broadcast and others. Conclusion Convergence is no longer something that we simply talk about. It is quickly becoming reality, and with convergence comes a greater urgency to meet customer demand in an expedited and seamless manner. New architectures and platforms such as M-CMTSs will provide the opportunity to quickly answer competitive threats, enabling the rapid introduction of new products and services, resulting in greater customer loyalty and increased ARPU. John Treece is director of Business Development, Cable Products Business, for Juniper Networks, and Michael Patrick is a data architect for Motorola. Reach them at email@example.com and firstname.lastname@example.org.