Splitting wood, splitting nodes. It’s a rough analogy, but helps get across several points.
First, both are established practices, and even popular, given cold temperatures or bandwidth constraints. Second, both cover a range of materials and techniques. Finally, as far as generating heat and bandwidth relief, both are limited in what they can achieve.
That last point makes this topic controversial. You can’t get too far into a discussion of node splits before hearing about alternatives. Big nodes, costly splits The idea of segmenting or splitting the fiber-optic node appeared soon after that device became the linchpin of cable’s HFC architecture in the mid-1990s. In cable years, that’s a long time ago.
The practice has accelerated in the past few years, and the near-term trend looks strong. “A lot of node segmentation is going to be performed in 2007,” says C-COR Vice President, Access Technology, Bill Dawson. Why is that?
One reason is that there remain some obvious targets. “We have customers who have currently 2,500-home node sizes now,” says Motorola’s Director of Systems Marketing, Connected Home Solutions, Fred Slowick. “That’s a very large serving area.”
Just how big? Slowick says most of the major MSOs are in the less-than-a-thousand homes-passed category. A commonly heard estimate is that most nodes serve between 500 and 750 homes passed.
Operators have commended this technique publicly. It is one of several reported “tools,” for instance, in Cox Communications’ extendable optical network (EON) initiative’s “tool kit.” Yet there’s a downside. Cox Senior DOCSIS Engineer Ben Bekele said in a recent CT-hosted Webcast on DOCSIS 3.0 that reducing node splits would be a good thing. Why? “They cost a lot of money.” Initial steps Through metrics such as Time Warner Cable’s Node Quality Index and Comcast’s 100-Worst and Most-Penetrated Nodes, operators pay close attention to the status and health of these devices. As for mapping them against available bandwidth, Charter Communications Vice President, Network Operations and Engineering Services, Keith Hayes underscores the importance of methodically gathering and acting upon the right kind of data.
“The first step is you need to know where you’re running out of bandwidth: whether it’s VOD (video on demand) or HSI (high-speed Internet) or telephony. No. 2, you need early warning.”
Before touching the outside plant, however, Hayes recommends optimizing signal transmission. That would include upgrading from lower modulation orders to 256-QAM (quadrature amplitude modulation) on the downstream and from quadrature phase shift keying (QPSK) to 16-QAM on the upstream.
Another technique is to rationalize narrowcast so that there is a 1:1 combining architecture between nodes and headend or hub optics. Given a three-way splitter servicing three nodes, for instance, the idea is to remove the splitter, add two transmitters and make those optical connections point-to-point.
Yet this practice faces its own alternatives. In the aforementioned Webcast, Comcast Senior Manager, Advanced Engineering, Saifur Rahman said that instead of “decombining,” operators should consider using DOCSIS 2.0 techniques to add more downstream DOCSIS RF channels per node. In either case, you’ve moved beyond modulation upgrades but not yet reached a node split.
“You want to take the easier steps, which are typically your denser modulations, before you take the physical steps of having to recombine or extend fiber and cut a node in half,” Hayes says. Knotty nodes Much as wood varies by species, fiber nodes come in all shapes and sizes, makes and models.
Segmentable nodes have been widely – though not universally – deployed. Yet even nodes designed to be split can end up serving an unbalanced, or asymmetrical, number of homes off each output port in the field, making them resistant – much like a piece of knotty wood – to clean strikes down the middle.
“I would say – just a guess – that 50 percent are segmentable,” says Motorola’s Slowick. “But 75 percent of all nodes out there are not balanced, so it doesn’t do you a lot of good.”
In those cases, the practice is not so much splitting a node as chopping away at a nodal service area. The idea is to pull fiber further down the line and install a smaller, satellite or mini-node in a subset of the original area. Of course, that adds costs.
In the best case, the coaxial plant is balanced and the node is either segmentable or can be replaced with one that supports the desired split. “It could be as simple as a one-hour, pull-out-old-node, activate-new-node,” says Charter’s Hayes. “You have somebody at the hub reconfigure the electronics and someone down in the field who removes the old node and splices in the new one.”
The “knotty” cases are more complicated. “You have to strive to accomplish some sort of a balance in the ultimate topology,” says Dawson. The need to realign the plant makes this a bigger project. “That might have to take place over a night or two previous to the actual node split,” says Hayes. “Instead of one technician, you bring in three or four to go do the reconfiguration.”
Lashing up express coax and installing mini-nodes farther down the line consumes additional labor and materials, including fiber. What fiber? Pulling a half-mile of fiber to a new node is not out of the question. But increasing the number of fibers feeding optical nodes is a nonstarter. Standard counts, ranging from eight to six to four (which becomes two within-a-ring in ring architecture) are fixed quantities.
Even operators who budgeted spare fibers in rebuilds a decade ago find themselves short today, having redeployed that glass to homes or perhaps businesses that popped up in the interim. “The point is, they are not there as spares anymore,” says Dawson.
That relative fiber scarcity has accelerated the number of wavelength division multiplexing (WDM) products in the access space. This trend has been building. Five years ago, Scientific-Atlanta announced it had mounted coarse WDM (CWDM) return path modules into optical nodes and deployed with a European operator. Three years ago, at the SCTE Conference on Emerging Technologies, Time Warner Cable’s Tom Staniec spoke enthusiastically about collaborating with a vendor about handling a “lambda on a pole.”
The category has matured, with operators apparently eager for more efficient use of their fiber asesets. “We’re getting a tremendous response for our forward path CWDM technology,” Dawson says.
The point is that plain segmentation itself only gets you so far. Adding optical muxing and de-muxing or mounting erbium doped fiber amplifers (EDFAs) in the field gets you further, but of course adds costs. WDM and more A recent development in this space is the adoption by Aurora Networks of dense WDM (DWDM)-related technologies associated with C band of the ITU grid (1,530 nm-1,565 nm) back into the O band (1,271 nm – 1,360 nm).
“We’ve been able to get to a point where we offer six wavelengths within that O-band window,” Aurora Vice President of Marketing John Dahlquist says.
Dubbed “LcWDM,” Aurora’s approach promises to facilitate node segmentation (with or without network reconfiguration) by delivering twice as many wavelengths over a single fiber and going two times the distance, from 15 km to 30 km, as CWDM. It also dovetails with Aurora’s established position on eliminating RF amplifiers, even down to a node-plus-zero architecture.
This architectural point is yet another way in which a discussion of nodes often segues into more strategic questions.
In a paper presented at last year’s Cable-Tec Expo, Eric Schweitzer (then Harmonic Senior Director of HFC Product Marketing, now with Comcast) tested the assumption that reducing the number of amplifiers has positive network operating results.
“The conventional wisdom turned out to be correct,” Schweitzer says. “The deeper you bring fiber into the network, the more reliable the network is.”
A live topic that Schweitzer also explored is what number of amplifier cascades will enable operators to support the CableLabs’ definition of required availability for VoIP networks (namely, 0.9994.) His answer? Node-plus-two with redundant power supply, node-plus-one without. Expansions, SDV How the industry eliminates enough actives to hit that target is a loaded question. (Schweitzer said the industry status quo was greater than node-plus-four.) More commonly, node-plus-n is discussed in terms of new build rather than existing plant.
Beyond the fiber-deep question, two other issues impact current discussion of node segmentation, namely expansion to 1 GHz and switched digital video.
As for expansion, it’s worth noting that today’s latest nodes already are supporting 1 GHz. The reason being, the optics cost the same. They’re also capable of mid-splits.
Whether, when and how widely any of that spectrum gets used is another matter. Some engineers on the operator side are beginning to use words such as “isolated single node, spot spectrum improvements,” carefully chosen to sidestep the trip wires that Wall Street has set to detect any potential movement toward network upgrades.
In the end, any analysis of node segmentation must answer that first question that Hayes raised: Where are you short? VOD? HSI? Telephony? And what about SDV?
Usually cast as a bandwidth savings technique, SDV also looks like a service, consumes bandwidth early on and reacts in its own way to a node split.
Whereas a split can double HSI capacity, “in the SDV case, there will not be a doubling,” says one MSO engineer, “because many program streams will be duplicated on each side of the split.”
“On the other hand, doubling the size of the narrowcast Edge (E)QAM allocation will provide SDV with far more than double capacity, because the additional EQAM resources will see mostly long-tail content.”
In other words, a split can work against SDV. “The efficiency gains you get in a 500- or 1,000-home node are much higher than you would get in a 250- or 125-home node,” says Jeff Fryling, vice president of business development at Vyyo, a firm that advocates operators’ cutting to the chase with bandwidth expansion, rather than getting hit with a “double spend.” Many weapons But in certain cases, segmentation may be compatible with SDV. That was one of the points that Dawson and C-COR colleague Jeff Sauter made in the July 2006 issue of CT. Although their article did not account for Edge QAMs, it did contain a busy table mapping existing against desired serving group sizes, indicating the kind of close calculations required to determine which techniques to use at which sized nodes.
The bottom line is that while large nodes with balanced distribution legs present easy targets, node splitting is no silver bullet in the battle for bandwidth. It’s one weapon in an arsenal that contains higher modulation orders, decombining, advanced DOCSIS and the latest environmentally hardened optics. Add to those the notion of spot spectrum improvement, boosts to Edge QAM resources and, needless to say, robust spreadsheeting capability. Jonathan Tombes edits Communications Technology. Reach him at email@example.com.
Splitting wood, splitting nodes. It’s a rough analogy, but helps get across several points.