“Pre-3,” “2.0-B, Whatever It Takes
by Leslie Ellis // April 24 2006
Talk about technology moving quickly. One of the bigger stories on the broadband side of the house got a predecessor, a new name, and then another new name — all within the last few weeks.
It all started during a pre-NCTA vendor briefing, when a maker of cable modem stuff dropped the term “pre-three” into a conversation. For context, here’s the exact sentence: “We think there will be significant demand for pre-three products — I don’t know if they’ll be able to wait.”
“They,” in this example, is cable operators. What they may not be able to wait for: Some of the goodies slated for the newest cable modem specification, known industrially as DOCSIS 3.0.
“Pre-three,” then, is shorthand for broadband-related gear that contains some of the features of DOCSIS 3.0, but not all of them.
Subhead: Meet “2.0 B”
Hang on. Don’t get too cozy with pre-three, because the name already changed — to “2.0 B.” (Marketers, before you freak, know these as industrial terms, necessary as internal guideposts to stuff that will help you compete.)
The “2.0″ in “2.0 B” refers to DOCSIS 2.0, the most recent version of the specification. The “B” stands for “bond” — channel, not James – one of the sexier ingredients of DOCSIS 3.0.
Why “2.0 B” and not “pre-three”? Because the cable operator side wanted to assure that manufacturers would keep going with all of the features of DOCSIS 3.0, and not stop at the early version. (I personally think “pre-three” is catchier than “two dot oh B,” but nobody asked.)
Why is this even happening? Because there are four giant features crammed into the full DOCSIS 3.0 spec. Finishing all four takes time.
The map from finished specification to actual product goes roughly like this: Silicon vendors “tape out” a chip design, based on their interpretation of the spec. That design goes to the foundry (the “fab”), which makes chips. Manufacturers buy the chips, add other necessary components, and build products. Those products go to CableLabs for compliance testing. From beginning to end, if nothing goes wrong, it’s 18 months to two years.
If the DOCSIS 3.0 specification were to come out today (all signs point to “imminent”), it’d be early 2008 before full DOCSIS 3.0 gear would be ready for market.
For many operators, that just isn’t fast enough. Operators competing internationally are up against 100 Mbps fiber connections; operators stateside have Verizon’s FiOS in their faces.
Subhead: What’s in 2.0 B?
Ultimately, the operators involved in building DOCSIS opted for a three-channel bond in the downstream direction, a one-channel upstream, and IPv6, as the core components of what is now “2.0 B.”
Reasoning: The full, DOCSIS 3.0 spec stipulates a minimum of four bonded channels, for 155 Mbps of downstream speed (4 channels x 38 Mbps). But operators don’t always have four unused digital channels available for the bond. Most of them don’t need that much firepower right away. Why not start at a three-channel bond, and bump up from there? That gets you more than 100 Mbps downstream.
IPv6 is in there because some operators are starting to worry (understatement) about running out of IP addresses for all the broadband gear they’re putting in people’s homes. Cable modems, voice-over-IP adaptors, and some advanced set-tops have IP addresses; some of those devices have more than one address.
IPv6 (translated in the Sept. 5, 2005 edition), offers a way to correct that. It takes today’s total address space, of 4 billion IP addresses, to a number so big, it doesn’t make sense to write it out. It’s a one with 18 zeros behind it. Everybody should be ok, as far as IP address space goes, once IPv6 is in place.
That leaves two features for the full, DOCSIS 3.0 specification: Advanced encryption, to even further secure consumer Internet connections, and a plumbing feature known as “M-CMTS,” for Modular Cable Modem Termination System (translated in the Feb. 28, 2005 edition).
Without re-hashing the whole thing again, M-CMTS makes it more cost effective to send video bits over the IP (cable modem) path, because right now it’s substantially more expensive than the digital set-top path.
Boil it all down, and it means this: With 2.0 B, operators get a fast track to better, faster broadband gear, and a fix to the IP address problem. Soon. Maybe even this year. Manufacturers get a faster track to revenues, making products based on 2.0 B.
All the while, everyone keeps pushing to the full specification, with forward compatibility likely to be enforced by purchase contracts. In short, everybody wins.
Switched Digital Part 2
by Leslie Ellis // April 10 2006
Last time, we looked at the many intersections between the technologies used for video on demand (VOD), and those used for switching video streams.
A brief review: Both work by setting up two-way “sessions” between a set-top and equipment higher up in the network–servers and streamers, in the case of VOD, and the switch, in the case of switched video.
Both also use the same transportation technique to move video streams to homes: Quadrature Amplitude Modulation, or QAM. So does “regular” digital video, HDTV, simulcasting, broadband Internet, and voice services. That’s why operators are maneuvering to make those QAM devices capable of mixing and matching any digital traffic that moves through them.
In short, VOD and switching tend to hang out together, conversationally, because they have overlapping features, and they can potentially share resources.
This time, we’ll look at what’s different about VOD and switching, and at what operators are doing to get ready for the switch.
Multicast, Unicast, and Switched Unicast
Usually, if you agree to go more than two or three mental steps down the path of learning about switched video, you’ll run into the term “multicast.” It is one of the chief differences between VOD and switched video.
Here’s how multicast works, by practical example: Customer Bob, there on the couch, lands on a channel he wants to watch. Let’s say it’s Oxygen (Bob being an evolved man, and all.) A session is set-up between Bob’s set-top, and the switch. At this point, only Bob is watching Oxygen, although it’s flowing to everyone in Bob’s node (usually a grouping of about 500 homes) and in Bob’s service group (a cluster of four or so nodes).
Next door, Customer Jane also tunes to Oxygen. Her set-top says to the switch, “she wants Oxygen.” The switch says, “it’s already flowing, tap in here,” and tells the set-top where to tune so that Jane can “join” that stream.
As soon as Jane joined the Oxygen stream, Bob’s unicast became Bob and Jane’s multicast. Ditto for anyone else served by that node, inside that serving area group, watching Oxygen.
But wait, there’s more: The notion of the “switched unicast,” which will be detailed in a technical paper presented by Big Band Networks at this week’s National Show in Atlanta. (The session is Sunday afternoon, but if you miss it, there’s always the compilation of NCTA Tech Papers.)
Here’s how switched unicast works: Customer Ray, down the street from Bob and Jane, tunes a different channel. Let’s say Ray wants the Food Network. In a switched unicast environment, Ray’s set-top asks the switch for that channel, and the switch responds with a unicast session, just to Ray. In a way, it’s like a VOD stream, but without the trick modes (fast forward, rewind, pause).
If Bob flips over to Food, he doesn’t join Ray’s stream, even though it’s flowing into the same node.
The advantage, switched unicast proponents say, is everything that comes with one-to-one personalization with customers the Ray-targeted ad, for instance. But, it takes more bandwidth than multicast, and it costs more.
Getting Ready for Switching
The week before last, our sister publication CED Magazine ran a webcast about video switching, where technologists from Cox and Time Warner Cable dispensed tips about how to get ready.
Among them: Allot plenty of get-ready time to rewire distribution hubs, which is where the switches often go. It’s the most time-consuming part of “going switched,” tech people say.
Also: Once and for all, get rid of any “non-responder” set-tops in the system. “Non-responders” are boxes that don’t or can’t communicate to the network, usually because they’re plugged into an outlet that gets turned off with the wall switch — or are otherwise “deaf” to the network.
Once the plant is ready, start slowly. Place “lightly viewed” networks on the switch first. Move only a few channels at a time, and determine what channels to move by using viewership analysis tools.
I’m not big into predictions, but, switched video bears a strong resemblance to “digital simulcast,” in terms of unanticipated priorities. Like digital simulcast, switching is emerging fairly suddenly on operator to-do lists, with aggressive launch targets over the next few years.
The reasoning: You don’t leave the lights on when you’re not home. You don’t leave the water running in the sink when you’re not brushing your teeth. So why put channels down the pipe that aren’t being watched?
I’ve yet to find one cable operator complaining of excess capacity. Switching conserves their precious digital bandwidth.