Cable Congress highlights importance of IP migration strategy for MSOs

European cable operators must be doing something right, if the rather bullish analyst predictions made during the recent Cable Congress 2011 in Lucerne, Switzerland, turn out to be true. But a closer look at the data reveals that these optimistic forecasts assume operators make the right calls in migrating toward all-IP fiber deep architectures where they will face increasing competition from IPTV and OTT services.

For now, European cable operators are enjoying a mini boom, according to research presented at the Cable Congress by Cable Europe, the cable industry's Brussels-based trade association, with total European revenue up 6 percent to €18.8 billion in 2010, double the 3 percent growth rate of 2009. The research, conducted by IHS Screen Digest, attributed much of the growth to the expanded services enabled by ongoing digitization, with a 26 percent rise in VOD revenue. With still only 41percent of homes digital, and ARPU at a relatively modest €23, Cable Europe sees plenty of scope for further growth.

But the recent growth spurt can only be sustained if operators maintain their current access bandwidth advantage over telcos, and this depends on successful migration to an all-fiber network. In the end, telcos and MSOs will be in the same place competing over a common service footprint with networks that are indistinguishable and may even share the same physical plant. But that is a decade or more away, and meanwhile, the fight will focus on content, ease of access, price of triple- or quad-play services, and headline bandwidth. At the moment, European MSOs are in the ascendancy, leading to this recent, largely unexpected growth because they, for the most part, have plenty of capacity in their HFC networks to exploit the channel bonding of DOCSIS 3.0 and offer 50 Mb/s or 100 Mb/s services. Some, such as Virgin Media in the UK, have also done well by importing the best of the Web into the living room with catch-up services and access to social media.

But the bandwidth gap will close as telcos push fiber deeper, and this was the subtext at the Cable Congress, with discussion around how best to migrate to all-IP architectures. One problem has been that the cable industry as a whole has had to shift its migration strategy several times over the last five years, finally ending up with a recent revision of the Common Multimedia Access Platform (CMAP), which was promoted originally by U.S. MSO Comcast, but now is supported enthusiastically by a number of European MSOs.

The first step on the road to IP came with the initial DOCSIS 1.0, designed at that stage purely as a way to deliver broadband Internet services over the RF HFC network, with broadcast services becoming digital but continuing to be delivered independently over the HFC via separate edge QAMs (eQAMs). QAM is the underlying scheme used by both DOCSIS and broadcast/multicast TV services to modulate signals, either digital or analog, over the HFC. At the early stage of DOCSIS, it looked like broadband Internet and TV services would remain separate, and that each subscriber would have just a single DOCSIS channel. The boxes were designed accordingly, with all functions, including timing, the eQAM and modem termination, integrated into a single unit. This had a number of advantages, including being relatively straightforward to deploy and maintain as an all-in-one unit and minimizing the number of points of failure. But as DOCSIS services grew and started to be used for unicast video as well, combined with the support for channel bonding in DOCSIS 3.0 to deliver higher bit rates to individual subscribers, it became clear that the integrated approach was insufficiently flexible, leading to the development of Modular CMTS (M-CMTS). This split the modem termination and eQAM functions into two so each could be scaled separately, with the latter deployed closer to the subscribers, enabling channels to be reallocated between them. The idea was that operators could make the most of their QAM investment without having to deploy more than was needed in aggregate for both DOCSIS and multicast/broadcast video.

But M-CMTS did not work as well as hoped, partly because of the complexity involved. Because the CMTS and eQAM were now physically separated, extremely accurate timing had to be maintained between them via a DOCSIS DTI timing reference. With M-CMTS, the eQAM and the CMTS can be remotely located, and multiple DTI timers can be used and synchronized via GPS. This allows the eQAMs to be deployed in hub sites where local video distribution is occurring, while the CMTS remains at the headend. In practice, where M-CMTS has been deployed, the eQAM has still been dedicated either to video or DOCSIS, rather than being shared between the two as had been intended, partly defeating the object because economies of scale are not then achieved.

In any event, a number of cable operators became disillusioned with the DOCSIS roadmap and took matters into their own hands. Comcast proposed a return to a single platform, CMAP, and this has now been supported across the whole industry. CMAP recombines eQAM and modem termination to reduce complexity, with fewer points of failure, avoiding the need for timing synchronization because the key components are once again in one chassis. The recombination of functions into a single platform also yields significant power savings.

CMAP does not have to be fully integrated and allows much greater flexibility than the original CMTS. Indeed, it separates the functions into two components, but these are the higher-level IP packet-processing functions, and the lower-level physical (PHY) and media access control (MAC) functions, rather than the eQAM on the HFC side and Ethernet on the headend or core network side, as with M-CMTS. The packet processing is provided by the so-called packet shelf, while PHY/MAC processing is provided by the access shelf. With this design, a CMAP platform can be integrated (I-CMAP), with the packet and access shelf combined, or modular (M-CMAP), with the two separate so they can be scaled independently. Even the integrated version is more flexible than M-CMTS because each port can modulate across the whole RF spectrum and support all services, including video, voice, broadband and IP unicast delivery, bringing much greater configuration flexibility without having to separate out the edge QAM.

In most cases, M-CMAP will be preferred to I-CMAP because it is more scalable, supports a pay-as-you-grow model and, in Europe, fits better with the distributed architecture many MSOs have, comprising a number of secondary headends fed from one single, large headend. The modular approach also allows MSOs to choose separate vendors for the IP and higher-level functions in the packet shelf and the MAC/PHY related tasks in the access shelf. This functional separation also broadens the overall pool of vendors participating in the CMAP arena, increasing competition and innovation. It also best serves the original motive for CMAP of providing a sound platform for the migration to all-fiber networks based on passive optical network (PON) technology, which could be deployed as a new access shelf rather changing the packet shelf. This also applies to other possible access options, such as wireless, which again could be deployed just by adding a suitable access shelf.

Yet the CMAP designers overlooked one important point, which is the CMTS legacy many operators were stuck with precisely because this was what the industry had been recommending before. A number of operators had taken the M-CMTS route, and CMAP as originally formulated could only be deployed by throwing this away. This did not go down well with many MSOs, and it threatened to derail the whole CMAP movement. Many of these operators had concentrated their upstream capacity into the CMTS using less expensive eQAM for downstream IP transmission, and they wanted to continue exploiting this investment.

The industry has just come up with a fix. This is to create a separate downstream-only version of the access shelf, which is in effect a super eQAM, creating the ability to hook existing eQAM up with the packet shelf, and so go on using it. This also required bringing back the CMTS timing mechanisms, DOCSIS Timing Interface (DTI) and Downstream External PHY Interface (DEPI). In an M-CMTS deployment, DTI keeps the clocks of the upstream and downstream components synchronised, while DEPI provides the IP link between the CMTS and the eQAM. DTI and DEPI will in the future be doing the same in a CMAP deployment where there is legacy eQAM to be supported. DTI and DEPI were not in the original CMAP specifications, however, because there was no significant physical separation between the access and packet-processing functions.

At least now, after a number of false starts, the cable industry appears to have stumbled upon a coherent migration strategy toward an all-IP world, while preserving legacy M-CMTS deployments.