As the cable industry prepares to deliver the next generation of broadband and video services, Light Reading has developed a four-part series of articles exploring the evolution of cable networks and the strategies that service providers are pursuing as they begin laying the architectural foundations for a vastly different future. Light Reading has looked at timing, technological and business considerations, and the advantages that software-defined networking (SDN) will ultimately bring. In this fourth and final installment today, we examine the early decisions the cable industry is facing as it transitions legacy infrastructure to new virtualized networking systems.

To catch up on the Light Reading series, click through for part one, part two and part three. Continue reading for part four below.

Place your bets
The cable industry recognizes the benefits of moving to a distributed and virtualized network architecture, but the paths any individual operator may take to get there are numerous and varied. Service providers aren't the only ones facing high stakes either. Vendors also have to decide where to invest their resources.

One of the big early debates is how to distribute functions in the cable network, and specifically, whether to distribute both the MAC and PHY layers of a cable modem termination system (CMTS) down to the fiber node, or only the PHY layer. (See Cable's Great Debate: How to Split Functions.)

The debate has been going on for several years, so much so that the industry is now actively trying to deflect discussion away from the clash of opinions and point instead to the fact that the end goal of both distribution strategies is the same. Whether an operator distributes just the PHY or both the MAC and PHY layers, the goal is to virtualize the management of network nodes and edge devices.

The problem is that by dismissing the MAC/PHY debate as a "religious war," industry experts may also be minimizing the decision that has to be made. Any new node is likely to sit out in the field for a decade or more, which means that choosing whether to remote just the PHY or the MAC and the PHY is a decision with long-term effects.

Of the primary cable vendors (not including Huawei Technologies Co. Ltd. , which has been largely exiled from the US networking market), only Nokia Corp. (NYSE: NOK) today is advocating for a Remote MAC/PHY solution (also known as Remote CCAP or virtualized CCAP). Part of Nokia's argument is that distributing both layers makes it easier to get rid of proprietary hardware in the headend because all of the cable-specific components are pushed out to the edge of the network. Nokia also argues that separating the MAC and PHY layers may create some timing issues, although the timing concern doesn't seem to be a major deterrent for others. (See Nokia Debuts Virtualized Cable Access Platform.)

Because it's the only company actively promoting Remote MAC/PHY in the US, Nokia looks uncomfortably like it may have gone out on a limb. But that doesn't mean the company is without supporters. WideOpenWest Holdings LLC (WOW) said in June that it was preparing to deploy the Nokia solution, and multiple cable leaders, including Comcast Corp. (Nasdaq: CMCSA, CMCSK) President of Technology and Product Tony Werner, praised Nokia's acquisition of the Gainspeed distributed access architecture (DAA) technology when that buyout occurred. (See Nokia Swings Deal for Gainspeed.)

Arris Group Inc. (Nasdaq: ARRS) CEO Bruce McClelland also makes it clear that Arris at least is not wedded to delivering only a Remote PHY solution. "If there's enough demand for remote CCAP, we'll do that," said McClelland at the SCTE Cable-Tec Expo show in September.

Video and virtualization
Another big decision cable operators have to make as they virtualize and distribute their networks is how to handle video in the near term. With both a Remote PHY and a Remote MAC/PHY solution, the link between a headend or hub site and a fiber node transitions from an analog connection to a digital one. That means cable companies need a new way to transport video to the node, at least until there's a full conversion to IP video delivery.

There are two primary options for handling legacy RF video transport in a distributed architecture. One is to deploy an integrated CCAP chassis that converges data and video delivery. The other is to keep data and video separate but convert RF video signals to digital format before they're transported to the node.

The advantage to the latter approach -- which is being promoted in different variations by Nokia and Harmonic Inc. (Nasdaq: HLIT) -- is that it allows operators to virtualize their data services more quickly and host them on commercial-off-the-shelf (COTS) hardware. However, the second approach is also the more radical of the two, and many operators are more comfortable taking advantage of integrated CCAP devices today as an incremental upgrade step.

For more fixed broadband market coverage and insights, check out our dedicated gigabit/broadband content channel here on Light Reading.

Vendors also have their own reasons for backing one strategy over another. Incumbents Arris and Cisco Systems Inc. (Nasdaq: CSCO) have every incentive to keep selling their expensive CCAP equipment even while they lay out a migration plan for eventual virtualization. Harmonic and Nokia, on the other hand, have a financial interest in moving customers off Arris and Cisco hardware and on to their own software-based platforms. (See How Harmonic Aims to Disrupt CCAP Market.)

Casa Systems Inc. is in a unique position. While the company is one of the leading CCAP suppliers alongside Arris and Cisco, it's also now saying that it can deliver integrated CCAP functionality in either hardware or software form. The jury's still out on how well the virtualized version of Casa's integrated CCAP works, but if cable operators find the solution viable, it could offer a third route forward for network migration. (See Casa Joins Virtual CCAP Parade.)

Whatever option, or more likely options, cable companies choose, the critical factor in analyzing success will be looking at the financial outcomes. Proprietary cable network hardware won't disappear overnight, so the question is how best to time the transition. It's likely that as operators exhaust the capacity of their integrated CCAP devices, they'll start adding off-the-shelf servers next to their CCAP hardware. However, that process could make for a very long migration timeframe, and there is a balance to be found between the gradual approach and one that gets operators to their desired end state more quickly.

So far, there's limited research on the ideal timeline for cable's network transformation, but that doesn't mean there's no information available. According to Justin Paul, head of OSS marketing at Amdocs Ltd. (NYSE: DOX), there's a lot of virtualization creeping into cable networks already, and he believes that cable infrastructure will be mostly virtualized within the next decade. However, from a business perspective, Paul also said at Light Reading's Virtualizing the Cable Architecture breakfast session at Cable-Tec Expo that ten years is probably too long for a cable operator to take once the decision to upgrade is made.

Five years for complete transformation presents a strong business case, said Paul. But, he continued, "if you take ten years, what you end up with is the challenge of managing your existing networks alongside virtual networks, and you don't get the cost benefits as quickly. So when you do make the decision to go, try and go as quickly as you can to reach that end point."

In other words, the transition phase is the hardest part of network virtualization, but at some point -- and within the next handful of years -- everyone in the cable industry will have to bite the bullet.

— Mari Silbey, Senior Editor, Cable/Video, Light Reading