By Phil Hunter
Remote PHY is emerging as the first-choice distributed access architecture (DAA) for many of the leading cable operators in all regions because of its combination of relative deployment simplicity and potential for significant cost reduction. The argument is not entirely settled however with some leading vendors such as Arris holding out for a choice of access architectures on the ground that the best option depends on factors on the ground, especially the operator’s legacy infrastructure. Meanwhile the field remains mired in confusion and misinformation, which is making it harder for smaller MSOs with less expertise to plot the right migration path towards an all-IP future where voice, data and video are all handled interchangeably over a common network fabric from the PHY upwards.
There are six choices of access architecture on the table, although some of these are centralized and one or two are gaining little traction. The tide is flowing towards distributed DAA with two main architectural contenders now generally recognized, Remote PHY (R-PHY) and remote MAC-PHY.
This is the latest and perhaps almost final chapter in the evolution of cable architectures ever since digitization began in the 1990s with the development of QAM for modulation of signals over the underlying HFC (Hybrid Fiber Coax) infrastructure. Then DOCSIS 1.0 emerged to enable transmission of IP data for Internet access alongside the video, but at first this required separate QAMs, as voice did as well. This was still the case when VoD emerged with one-to-one unicast delivery of video, as well as Switched Digital Video (SDV) to switch QAM video channels at the fiber node as a form of multicast to free up bandwidth.
But then around 2007 the edge QAM was developed to integrate voice, voice and data QAMs in a single multi-purpose platform. Soon after that CCAP (Converged Cable Access Platform) emerged in turn to combine the edge QAM with the central cable modem termination system (CMTS) functions into a single super-dense platform as the next step towards convergence around an IP video platform. This opened the door to virtualization of the central network up to the fiber node and use of COTS (Commercial Off the Shelf) hardware.
This set the stage for the era of cable IP migration, which then heated up with the release of DOCSIS 3.1 in 2013. This at last combined the longstanding two versions DOCSIS and EuroDOCSIS into a single worldwide standard by ending the 6 MHz and 8 MHz wide channel spacing used respectively by the two versions and replacing that with OFDM (orthogonal frequency-division multiplexing). CCAP was then the platform for converging DOCSIS IP and QAM based video delivery within a single hardware platform. It unites DOCSIS and video QAM traffic within a single RF connection so that the two can be switched purely in software.
The node could then be either R-PHY or Remote MAC-PHY. R-PHY distributes the PHY function to the fiber nodes while retaining central control over the MAC, while MAC-PHY distributes both. To assess the merits of each we have to understand that on a cable network the MAC layer does more than its original intention as a layer two protocol just framing data for transmission at the link level between two nodes or around a local area network as in Ethernet.
Because DOCSIS was designed as an integrated system to transmit data on an end to end basis across the HFC network from CMTS in the core to the cable modem in the premise, it also incorporates some functions normally associated with layer three at the networking level, such as being able to classify and prioritize IP packets. Under MAC-PHY, the distinction between layer two and layer three as in IP/Ethernet networks is restored, so that those more complex features are retained centrally while the traditional layer two functions are distributed to fiber nodes. In the case of R-PHY, all MAC functions are retained centrally while just the PHY is distributed. The advantage of MAC-PHY is that the centralized systems are simpler, but this is more than outweighed by the complexity of splitting the software into two pieces and requiring extra resources on the distributed R-PHY nodes to execute even the basic MAC functions. At least that is the view that most operators, as well as Cisco, have been tending towards.
Liberty Global’s CTO and SVP Balan Nair indicated recently that he was leaning towards R-PHY and away from Remote MAC-PHY for these reasons as it proceeds with its DOCSIS 3.1 roll out, while in the US Comcast and Cox Communications have both made noises in this direction on various occasions. Significantly Liberty Global and Comcast will be leading a “deep dive” into R-PHY at the SCTE-ISBE Cable-Tec Expo 2017 this month.
Meanwhile the list of vendors offering R-PHY products continues to expand. The big players, including Cisco, Harmonic and Arris are all there, while Nokia has become a major force in cable through its acquisition of Alcatel-Lucent and then Gainspeed, a US start-up specializing in virtualized CCAP.
As the field ramps up smaller vendors are coming in, such as Teleste, a $240 million annual revenue Finnish broadband products firm, with a focus on remote monitoring and automated set up. Its R-PHY fiber node can act as a monitoring probe for the network as a whole, locating failures and also gathering data about network and device status. The idea is that the cost savings achieved by having a relatively simple node running just the PHY should be complemented by opex savings.
According to Teleste, the main point about R-PHY is that apart from helping move to a virtualized IP infrastructure it also dovetails with pushing fiber deeper to increase capacity as the coax segments are shortened. By distributing the PHY, RF modulation can be replaced by IP/Ethernet as far as the node. The links from the central CCAP to the node could be Ethernet PON (EPON), Gigabit-capable Passive Optical Networks (GPON), or Metro Ethernet (MetroE), in each case offering cost savings as well as capacity increases, while supporting all DOCSIS services with end to end QoS assurance.
These Ethernet architectures mean that the cable-specific portions of the infrastructure now only begin at the node and that from there upstream there is no difference from Telco networks. This point has not been lost on some vendors such as Cisco which are promoting R-PHY as an in-building solution to Telcos as well as MSOs. We have not heard of any major Telcos taking this idea up yet, but it could be a good way of competing in the MDU (Multi Dwelling Unit) market in those countries such as the US and some in Europe where such buildings are quite widely cabled with coax. By running fiber to the building they could provide gigabit services to the unit.
However we should not run ahead of ourselves. The reality, as spelt out by Liberty Global’s Nair, is that most substantial cable networks have a lot of CPE that is still not ready for DOCSIS 3.1 and therefore for R-PHY. Liberty Global itself is waiting for the price of integrated DOCSIS 3.1 gateways to come down below $100 before starting a full-blown roll out. The chips are the cost bottleneck since they have to support WiFi as well and of course MoCA if there is going to be in-building network support in those MDUs. It is also worth noting that these emerging architectures are supposed to support enterprise applications as well as serve public WiFi hot spots so that MSOs can extract maximum revenue from their networks.
At present therefore Liberty Global is confining its DOCSIS 3.1 roll out to around 10% of its subscriber base this year before embarking on a full scale second phase in 2018. That is the situation most MSOs find themselves in.