Greetings, colleagues, and especially those of you headed to the ANGA conference in Cologne next week. I’m expecting there will be considerable discussion about current events in broadband, and as such thought I’d impart my latest thinking on one of my favorite topics – the Data Over Cable Service Interface, or DOCSIS. (It’s my favorite because I spend most of my waking hours thinking about broadband architectures and implications, and DOCSIS tends to be a big part of that! The latest specification is 947 pages long…)

By now, D3.1 and its successor, D3.1 FDX (Full Duplex) are as much a part of the industry’s parlance as were its previous chapters, which go back nearly a quarter of a century. In the beginning (’97-’98), we were focused on how to get Megabits on the wire – 27 Mbps down, specifically, and 2.5 Mbps up. (The former was a function of QAM modulation; the latter of QPSK.) That was DOCSIS 1.0. It was happening at a time when the only alternative was the dial-up telephone modem, with max connect speeds measured in hissy, scratchy 28 or 56 kilobits per second/kbps.

DOCSIS 1.1 brought the QoS features that begat VoIP; D2.0 brought the modulation advancement (ATDMA) that begat channel hopping. Then D3.0 introduced channel bonding, and IPv6 to address what was then a big concern about the IPv4 address space vanishing. In parallel with D3.0 came the DOCSIS Set-top Gateway (DSG), modular CMTS devices and the Remote PHY steps that positioned the world’s broadband networks from centralized to decentralized; from proprietary to standardized; from fixed to routed.

Which brings me to Musing #1: In the future, every node will connect to every hub site – and the spectrum at the hubs will be generated by a chip in a node.

The fact that service providers are contemplating or in production with a conversion of their HFC infrastructure from analog to digital and from fixed to routed is, in and of itself, huge. Transformative-huge. Here’s what I mean by that: Today, in any HFC network, its engineer could tell you which node is connected to which hub site. It’s a hierarchical topology, easy to draw on a napkin: There’s a combining network that takes the 150 or so coaxial outputs into outbound spectrum, which gets pumped into the optical network and sent out on the plant. In the future, the spectrum gets generated in an ASIC, inside a node. That matters because the hub site will have to find a new role as an Ethernet switching center along with servers for edge compute – but one day there will be no RF in the hub.

When paired with automation and orchestration techniques common in “the cloud,” it makes turning up a node happen in the time it takes to gaff a pole – versus a half a day or more. Node-based spectrum generation, with all nodes aware of each other and connected – it’s a mind-blowing architectural advancement with implications that’ll take some time to fully unpack.

Musing #2: DOCSIS 3.1 will give service providers at least a decade of growing room.

DOCSIS 3.1 led with the Gig. That was the big thing, in the beginning: Enabling service providers to offer competitive, reliable, Gigabit speeds, down the network. These days, three other “pillars” are getting more visibility as part of the motherlode that is D3.1, namely, latency (lower is better), security (higher is better) and reliability (highest is better.) Beyond that is D3.1 FDX, which packs even more architectural firepower, both by leveraging Distributed Access Architectures. FDX started out in the DOCSIS specification as an amplifier-less infrastructure (aka “N+0”), but we think we can take FDX to N+2 or even N+4. FDX initially provides 50 times the upstream bandwidth compared to today’s 42 or 65 MHz return path, and does so by sharing the same spectrum as the forward path. Now that is cool. Meanwhile, the forward path is starting to get a roadmap to 1.8 GHz and above. (I think we could get to 3 GHz, but that’s another blog.) The point of FDX is the duplex part: D3.1 aims, ultimately, for 10 Gigs down and 10 Gigs up. Again, I’m talking “in the fullness of time” here, and about how D3.1 and D3.1 FDX give operators plenty of growing room.

As you might imagine, I have plenty more musings on this topic! If you’ll be at ANGA, I’ll be on a panel about this and other topics at 11A on June 4, in Room 1. Come find me!



John Chapman

Cisco CTO Broadband & Fellow

Cisco’s Cable Access Unit