Platforms on Offer from Casa, Harmonic, Huawei and Nokia Raise SDN vs. NFV Question
By Fred Dawson
October 29, 2016 – After a protracted multi-year debate over best approaches to CCAP virtualization Cable MSOs finally have some real product options to consider as they look for ways to maximize the power of DOCSIS 3.1 to meet bandwidth and service requirements in the years ahead.
The aim of these vendor initiatives is to capitalize on the virtualization efficiencies made possible in CCAPs (Converged Cable Access Platforms) optimized to work with the CableLabs-defined Distributed Access Architecture (DAA) model, which relies on digital optics terminated at the HFC node by Remote PHY electronics that manage modulation, multiplexing, forward error correction and other physical layer processes in the conversion to RF for distribution over coax. The question posed by the new vendor options is, once the CCAP is relieved of performing these PHY layer processes, what components of the DAA–enabled CCAP should be virtualized.
In one approach, as exemplified by Casa Systems’ recent demonstration of its forthcoming vCCAP solutions at Cable-Tec Expo in Philadelphia, both the control and data plane components of the CCAP are virtualized to run on COTS (commodity off-the-shelf) servers at headends and datacenters. Harmonic, too, has introduced a fully virtualized CCAP dubbed “CableOS.”
The other approach, known as a Remote MAC-PHY configuration, is supported by new products from Nokia and Huawei. Their solutions virtualize the control plane functions running in core locations to orchestrate provisioning, quality control and tie-ins with other back-office elements but move the CCAP data plane or MAC (Media Access Control) into the node.
From a CableLabs’ specifications standpoint there’s a complex set of options associated with Remote MAC-PHY having to do with how much of the MAC resides in the Remote MAC-PHY Device (RMD) and how much remains in the hub or headend. The Nokia and Huawei solutions entail the most comprehensive encapsulation of the MAC in the RMD, which is referred to in the specifications as Remote CCAP.
Now that real products are entering the market, it will be interesting to see how operators respond. It’s not an easy choice, given the far-reaching implications of the pros and cons associated with each approach and how they play into DAA.For Nokia the appeal of the Remote MAC-PHY approach was strong enough to merit acquisition of Gainspeed, which provided the CCAP functions to complement the Nokia-supplied digital optics and Remote PHY components. “By co-locating the CCAP MAC with the Remote PHY electronics in our new SC-2D node operators can reduce space consumed in the headend compared to the other Remote PHY approach by a factor of seven,” says Stefaan Vanhastel, head of fixed networks marketing at Nokia.
Modeling how its solution would benefit an unnamed Tier 1 MSO in a move from a non-virtualized CCAP environment to DAA, Nokia found the platform delivered an eight-fold reduction in power, a seven-fold reduction in rack space and a substantial improvement in signal quality. And like any Remote PHY solution, the MAC-PHY approach eliminates the fiber link distance and wavelength density limitations imposed by analog optics.
Erick Keith, principal analyst for broadband networks and multiplay services at Current Analysis, offers an even more upbeat assessment. The Nokia solution, he says, “takes MSOs to the next level with at least three major ‘Force 10’ efficiency multipliers – specifically, 10x improvements in fiber efficiency, power consumption and rack space footprints over centralized CCAP implementations.”
But Casa believes the industry consensus is settling around another view. Loading up nodes with electronics adds headaches that operators don’t need, especially when the ongoing density gains of commodity processors in combination with other efficiencies tied to NFV (network function virtualization) architecture promise to greatly alleviate the space and power burden in headends, says Matt Eucker, manager for U.S. sales engineering at Casa.
“With Remote MAC-PHY you’re essentially hanging a CMTS on a pole, which leaves you with potentially hundreds of CMTSs to manage from the headend,” Euker says. Adds Casa CTO Weidong Chen: “Anybody can build a remote MAC. We just don’t feel that’s the right architecture.”
Whichever way operators choose to go, there’s little doubt the industry is on the cusp of a major transition to distributed access architecture (DAA), with or without CCAP virtualization, A year ago, IHS-Infonetics, reporting on in-depth interviews with cable companies that collectively control 87 percent of the industry’s capex, found that 42 percent planned to deploy DAA in at least some facilities by 2017.
The adoption of DAA is spurred in part by strategies aimed at maximizing the benefits of DOCSIS 3.1, which survey respondents on average said would reach a third of their subscribers next year. DOCSIS 3.1, with use of much higher orders of QAM modulation, improved forward error correction techniques and the multiplexing efficiencies of OFDM (orthogonal frequency division multiplexing), is designed to utilize up to 1.2 GHz of RF spectrum on the coaxial links to support up to 10 gigabit-per-second speeds downstream and 1 Gbps upstream.
DAA makes it easier to push fiber deeper, not only to be able to reach the 1.2 GHz level of usable coax spectrum with elimination of long amplifier cascades, but also to continue reducing service group sizes once the 1.2 GHz capacity is reached in order to raise the amount of throughput available to each subscribing household. DAA offers the same fiber migration benefits in conjunction with use of DOCSIS 3.0 as well.
Remote PHY is indifferent to what type of approach is taken to implementing distributed CCAP technology. It works with CCAPs running on proprietary hardware as well as virtualized CCAPs running on commodity off-the-shelf (COTS) servers. But with DAA expediting migration to ever more nodes, it appears that CCAP virtualization will eventually be required to more efficiently accommodate the multi-service needs of a growing number of service groups.
When it comes to assessing which approach to virtualization is better, one big question is what, if any efficiency gap exists between the Remote PHY and Remote MAC-PHY models. Claims from vendors taking either approach are equally impressive.
For example, Harmonic says CableOS used in conjunction with DAA saves up to 90 percent on space and power costs. Where a hardware-based CCAP typically requires nine racks of equipment to support 80 service groups, CableOS running on four racks can support more than 250. With these kinds of gains, the efficiency tradeoff question comes down to whether operators believe further gains are worth pursuing with remote placement of the MACs.
It’s also important to recognize that both approaches can claim the benefits of using digital optics with Ethernet transport. Along with the ability to support denser wavelength packing over each strand of fiber at much greater transmission distances than is possible with analog optics, digital produces a higher carrier-to-noise (CNR) output after conversion to RF at the node, which in many cases enables use of higher orders of QAM on the coax.
Huawei, for example, notes that use of digital fiber connected to one of its Remote PHY nodes allows service providers to achieve the 41 dB CNR, which is the minimum RF output at the node required to support 4K QAM with OFDM on DOCSIS 3.1 connections. Looking at current node positioning, this level of modulation supports an average 8x gain in the number of people in the service group that can be served at 1 Gbps, Huawei says.
“Our Distributed Cable Converged Access Platform is a fully baked product that now supports DOCSIS 3.1,” says Sean Long, director of optical network product management for North America. (Huawei’s first D-CCAP release only supported DOCSIS 3.0.) “We’ve announced deployments in Denmark and Peru,” Long says. “And we have some engagements with smaller operators in the U.S.”
Nokia, too, reports strong interest in its Gainspeed Virtual CCAP (V-CCAP) portfolio with a number of lab and field trials underway across North America and Europe in advance of a commercial release scheduled by year’s end. By early 2017 Nokia will augment the options with release of a Gainspeed node supporting 10 Gbps downstream throughput on DOCSIS 3.1 links. And, as previously reported, Nokia’s Bell Labs is working on full-duplex 10 Gbps DOCSIS 3.1 (FDX), a still developing protocol that utilizes full 1.2 GHz of coax spectrum without any amplifiers on the coax link.
Nokia’s confidence in the Remote MAC-PHY approach to DAA rests on the notion that operators will want the flexibility to manage service offerings and spectrum allocations independently within each service group and to have more options when it comes to future migration of nodes and access technology. “The ability to remotely configure and manage each access node will be increasingly important to operational agility in the years ahead,” Vanhastel says.
This includes the ability to make the transition from legacy video to IP video on a node-by-node basis. With introduction of the V-CCAP solution Nokia is making use of the Gainspeed Video Engine, which is designed to protect operators’ deployed Edge QAM (EQAM) investments by terminating RF video services in the hub, thereby enabling Ethernet over digital optics transport to the Gainspeed nodes where the legacy video service is delivered over the existing QAM infrastructure. This allows operators to begin offering an all-IP video service over the DOCSIS 3.1 spectrum without affecting subscribers to the legacy service while simplifying eventual transition of the entire node service area to IP-based TV service.
Another important aspect to migration agility is the support Nokia’s unified cable access solution can provide for PON, point-to-point Ethernet and Wi-Fi access points. Currently, Vanhastel notes, Nokia’s EPON node can be positioned next to or in place of the Gainspeed nodes to share the existing digital fiber resources in support of fiber extensions to customer premises and public Wi-Fi APs. “In the future we’re looking for ways to combine our PON and V-CCAP nodes into a single node,” he adds.
Another future development path involves use of the HFC plant to support 5G wireless access, which, operating from small cell locations, will enable 1 Gbps or higher throughput over very high frequency millimeter wave spectrum to small clusters of end users, The environmentally hardened nodes packed with miniaturized electronic modules might eventually be used as points of fiber extension to multiple small cells, allowing the backhaul support for 5G services to be provided over the HFC digital fiber infrastructure, Vanhastel explains.
Nokia is already looking into ways to configure backhaul architecture to maximize efficiencies, he adds. On the one hand, cloud-based management of the small cell infrastructure is vital to scalability, but transport of all the raw RF feeds from clusters of small cells would require “ridiculous amounts of bandwidth,” he says.
The solution may lie with a “mid-haul” approach that does some processing to consolidate the raw feeds at some point in the network while leaving it to the cloud platform to do the heaviest lifting. “We’re looking at eight or ten mid-haul scenarios, Vanhastel says.
In pursuing the Remote MAC-PHY path to CCAP virtualization, Nokia is utilizing the approach to virtualization known as SDN (software-defined networking), where a centralized control plane manages the data planes embedded in field-deployed purpose-built appliances. The Nokia Gainspeed Access Controller is the brains of the V-CCAP solution enabling operators to configure and manage a large and widely deployed network of MAC-equipped access nodes.
This is in contrast to NFV architecture where both the control and data planes are virtualized, doing away with dedicated purpose-built appliances altogether. “Virtualizing everything is not that easy,” Vanastel says. “We’ve taken a pragmatic approach using virtualization in whatever ways make the best sense for solving actual problems. If you want to maximize network migration agility, you still need physical interfaces in the access network.”
Casa’s Chen sees things differently. “We believe the next step in CCAP virtualization is NFV,” he says. Coding used with Casa’s vCCAP software is agnostic to whether the NFV architecture is based on OpenStack, containers or other emerging technologies, he notes, adding, “NFV allows you to scale much more easily.”
As described by Matt Eucker, the Casa solution maximizes scalability by enabling separation of the virtualized control plane and MAC functions so that one core datacenter-positioned control layer can manage virtualized MAC software positioned in multiple headend or hub locations to control service flows across a large number of nodes. A key advantage of this approach to distributed MAC management is it “closes the timing loop” that might cause unacceptable latency between the control plane and MACs if the MACs were positioned farther away in every node.
Adding to Casa’s case for a pure NFV solution is the fact that the vCCAP solution plays into the larger cloud framework of its Axyom software architecture, Eucker notes. Axyom is an open-source virtual edge mobile computing platform that provides a common secure foundation for independent scaling of control and data plane functions. With modular components supporting Wi-Fi wireless access gateway, LTE, 5G and other functions, Axyom establishes an NFV environment for vCCAP that enables dynamic use of COTS resources across multiple virtualized applications.
Clearly, MSOs face some tough choices as they move to utilize virtualization technology in conjunction with implementing DAA across their DOCSIS 3.1 footprints. Equally clearly, it looks like they’ll have the ability to ride virtualized CCAP technology a long way into the future, whichever path they choose.