Heading for DOCSIS 3.0
The industry drivers for next-generation DOCSIS systems initially included lower equipment costs and a competitive response to fiber to the home/business (FTTx). The precepts were that cable operators could leverage the cost curve of low-cost edge quadrature amplitude modulation (QAM) modulators used for video services, and that this low-cost capacity, coupled with channel bonding, would address competitive threats like FTTx.
To this end, CableLabs created the DOCSIS 3.0 specifications. DOCSIS 3.0 will provide operators with the high bit rate services needed for revenue protection and growth in residential and commercial services. It will also provide an architectural framework to support Internet protocol version 6 (IPv6) and allow traditional MPEG video and IP/DOCSIS services to share common network elements.
The industry has constantly refined its initial goals of capital savings and the delivery of high bit rate services. Additional DOCSIS 3.0 business drivers include investment protection, operational simplicity and RF spectrum efficiency. Investment protection reduces costs by leveraging existing cable modem termination system (CMTS) components for DOCSIS 3.0. Operators are acutely aware that reduced complexity among these systems will absolutely reduce mean time to diagnose/repair (MTTD/MTTR) and provide operational simplicty. These are critical factors for cable operators when deploying telephony and high bit rate services where service level agreements (SLAs) need to be maintained; however, operational support costs for all services benefit from solutions architected for high availability and reduced complexity. Lastly, services such as high definition TV (HDTV) and video on demand (VOD) are generating large amounts of revenue. These services along with “must carries” and the introduction of digital simulcast are increasing spectrum consumption. Therefore, spectrum efficiency is highly valued for both pre-DOCSIS 3.0 and fully compliant systems.
Increased competition has led to a surge in demand for DOCSIS 3.0. This has led vendors to develop pre-DOCSIS 3.0 systems before fully DOCSIS 3.0 compliant systems are available. Cable operators are interested in DOCSIS 3.0 technology today, with channel bonding in both downstream and upstream directions. Although the take rate for such services may initially be low, many cable operators are actively looking for timely vehicles to provide:
• High bit rate data services to compete with the residential threats of FTTx;
• High bit rate data services to compete in the commercial services market;
• Services to fill a void between current DOCSIS-delivered solutions and those delivered by fiber. Equipment cost drivers People sometimes ask, “Why does a CMTS cost so much when compared to other technologies such as Ethernet?” The answer lies in the DOCSIS media access control (MAC) methodology and its ability to deliver reliable per service flow quality of service (QoS). The DOCSIS MAC requires the use of an intelligent centralized scheduler, which necessitates computation. Contrast this with Ethernet’s MAC, carrier sense multiple access with collision detection (CSMA/CD), which has no scheduler and limited intelligence.
The CSMA/CD protocol is simple and low-cost; however, CSMA/CD cannot guarantee QoS. Mechanisms to add QoS to Ethernet such as type of services (ToS) or differentiated services code point (DSCP) do not come into play until after a packet has been transmitted. Ethernet has circumvented this lack of QoS by abandoning CSMA/CD for switching. Since each subscriber is isolated to a separate twisted-pair or fiber, switching transforms Ethernet from a shared media technology to a point-to-point technology. In addition to its lack of QoS, the performance of CSMA/CD over the required DOCSIS network distance of 100 miles is abysmal. If CableLabs had chosen a CSMA/CD MAC for DOCSIS, cable operators would never be able to achieve their desired throughputs. The intelligent scheduler of the DOCSIS MAC makes both QoS and performance possible.
Beyond the complexities of its scheduler, a DOCSIS switch, unlike an Ethernet switch, stores several state-variables per service flow needed to shape traffic, to enforce unique SLAs and to uniquely encrypt traffic. Each of these per flow capabilities requires silicon as central processing unit (CPU) gates, memory gates, application-specific integrated circuit (ASIC) gates, or field programmable gate array (FPGA) gates. An average DOCSIS CMTS has more capabilities per flow than most Ethernet switches have per port. Essentially, the CMTS can be compared to an Ethernet switch with more than 500,000 ports, where all data streams can be prioritized, shaped, rate limited and encrypted independently.
So how does one make DOCSIS technology cheaper while at the same time maintaining its exceptional functionality and performance? The answer is the same as it has always been—take advantage of Moore’s Law and increase density. Increases to silicon densities in the CMTS and QAM channel densities for a single upconverter will increase the total CMTS channel density and reduce per-port cost. At the same time, the greater densities will allow for more amortization of common components in the CMTS and even further reduce per-port costs. Considerations for channel bonding The DOCSIS 2.0 specification did not include increases in downstream data rates, so work on addressing higher downstream throughput began before work on DOCSIS 3.0 started at CableLabs. Increasing the downstream modulation from 256- to 1,024-QAM and/or increasing the channel bandwidth to 12 or 16 MHz were some of the alternatives considered. Ultimately, CableLabs chose packet-based channel bonding as the desired technique for increasing throughput.
Among the criteria used for technology selection were performance, cost and coexistence. Performance and cost are self-explanatory. Coexistence consists of two main principles. The first principle is the ability for legacy DOCSIS 1.x and 2.0 modems to coexist on all downstreams in the bonding group. This reduces the total number of DOCSIS channels required per fiber node by not requiring dedicated legacy channels or dedicated wideband channels. The second principle is that bonding may occur simultaneously on all downstreams; thus, a dedicated downstream for signaling would not be required. The lack of a signaling channel reduces the total number of dedicated DOCSIS channels required to achieve a desired maximum data rate. Coexistence has the added benefit in that the cable operator can achieve statistical multiplexing gains for the legacy subscribers by enabling dynamic load balancing in the CMTS. This would eliminate the need for very expensive node splitting in many cases.
CMTS architectures that are not fully converged will require an overlay scenario, dedicated RF spectrum supporting only channel-bonded customer premises equipment (CPE), forcing the operator to consume additional RF spectrum to deliver the same level of service. (See Figure 1.) Therefore, utilizing the same RF channels supporting legacy DOCSIS devices and pre-DOCSIS channel bonded equipment requires the least amount of RF spectrum consumption and the lowest total cost to enable high bit rate services. Cable operators should consider all of these criteria if they are planning to deploy a pre-DOCSIS 3.0 channel bonding solution. CMTS(s) to support 3.0 Selecting a CMTS with the proper set of features provides any cable operator with a great start toward a migration to full DOCSIS 3.0. When selecting a CMTS architecture, consider adopting a three-pillar approach: performance, availability and scalability. The selected CMTS architecture should allow an operator to achieve competitive DOCSIS 3.0 features in the future, such as downstream and upstream channel bonding, while retaining a vast majority of the capital investment. The changing competitive landscape requires cable operators to respond quickly to new and rapidly evolving threats from digital subscriber line (DSL) and FTTx providers without having to replace all or most of the existing CMTS components.
CableLabs specifications support both integrated and modular CMTS architectures. One approach is called modular CMTS (M-CMTS). The M-CMTS allows for the separation of the MAC and physical layer (PHY) functionalities into separate devices, requiring the creation of external communication protocols between these elements. These communication protocols between the elements are currently at various stages of CableLabs development.
The integrated CMTS (I-CMTS) architecture permits all MAC and PHY functions to reside in a single platform with no dependencies on external communication protocols. It allows for variable ratios of upstream to downstream channels and cards, while continuing to preserve high-availability protection schemes. (See Figure 2.) The selection of a CMTS must include an architecture that allows maximum configurability to enable any network architectures of the future. It should allow configurations where all downstreams originate from within the CMTS chassis via highly dense downstream modules (I-CMTS) and also allow configurations where all downstreams can originate from an external M-CMTS edge QAM modulator. CMTS architectures should offer cable operators the ability to increase density inside the existing chassis. This can be achieved by condensing functions onto fewer modules and allow more upstreams and downstreams to be supported in the same chassis.
The three-pillar approach when selecting any CMTS architecture (performance, availability and scalability) should also apply. Regardless of which path (I-CMTS or M-CMTS) is selected when migrating toward DOCSIS 3.0, the cable operator should still be able to depend on that CMTS for continued reliability and capital investment protection for years to come. Checklist for migration The competitive threats and revenue generation opportunities for residential and commercial services are driving pre-DOCSIS 3.0 solutions, and while CableLabs DOCSIS 3.0 specification development continues, operators are moving forward. In servicing commercial customers, high bit rate service may require costly fiber builds; however, with pre-DOCSIS 3.0 channel bonded approaches in both downstream and upstream directions, symmetrical services are now possible, thus avoiding costly fiber builds to meet return on investment (ROI) targets.
In addition to revenue generation, there are several other key factors cable operators should consider while planning for pre-DOCSIS 3.0 deployments and the migration to full 3.0. The CMTS architecture chosen should ensure investment protection for existing systems, decreased future capital expenses (per Mbps), flexible CMTS architecture paths (I-CMTS and M-CMTS), and minimized system complexity to reduce operational impact.
When deploying pre-DOCSIS 3.0, another key factor is the consumption of RF spectrum and coexistence with legacy DOCSIS devices. It is critical that pre-DOCSIS 3.0 channel bonded products coexist with legacy DOCSIS devices and share the same RF DOCSIS channels for management and data traffic. The allocation of dedicated standalone RF spectrum for channel bonded devices will drive superfluous capital and opportunity costs for the cable operator.
Among cable operators, as well as within each cable company, there may be different operating environments, products offered, and service level expectations, which formulate into architecture requirements. Consequently, it is critical that both pre-DOCSIS 3.0 and the migration to fully compliant DOCSIS 3.0 solutions promise and deliver flexible CMTS architecture paths to meet the needs of each operator’s environment(s) and the service levels expected by the paying customers, the shareholders and the cable industry as a whole. Mike Emmendorfer is senior director, Solution Architecture and Strategy, Office of the CTO – Broadband Division; Steven Krapp is director, C4 CMTS Product Management; and Brian Wheeler is senior manager, C4 CMTS Product Management, all for Arris. Reach them at [email protected], [email protected] and [email protected].
