"Broadband" Misconceptions and the (Advent) Everything over IP Future
By Steve Rose
Thursday, October 24, 2002
Broadband deployment in the United States has been slow, and a lot of misinformation has surrounded its deployment.
First misunderstanding: Cable modem or DSL service costs a lot more than dialup. Talk about a poor marketing job! Peak Internet usage corresponds to prime time -- when people are home from work, in the evening, the same times that peak telephone usage occurs. Dialup implies a significant portion of a single line telephone bill represents value being used for Internet access, or it may be a direct cost of a second line. Either way, the cost of the dialup line is $10 to $20 per month. Added to that is the cost of Internet access from the ISP, which varies from $10 to $24 per month. So the real cost of a dialup connection ranges from $20 to $44 per month (without an allowance for wasted time of the user while the modem connects or reconnects, or time spend waiting for slow data transfers, or the opportunity cost of missed calls). On the other hand, the cost of cable modem or DSL service varies from $25 to $50 per month, and offers five (50K bits vs. 250Kbits) to one hundred (25K bits vs. 2.5 Mbits) the access speed for the same (bursty) data, as defined above. In other words, a Charter cable modem connection at $25 per month saves money over a typical dialup, and even DSL at $50 has only a $10 to $15 real cost for an order of magnitude more speed.
Second misunderstanding: Although one can measure throughput of around 2 Mb/sec with RoadRunner (for example, based on the average of DSLReports.com speed test results), with peaks up to 4 Mb/s, this is based on two things: the duration of the test (about 30 seconds), and the current usage profiles. The duration has two effects: By averaging over an interval, the second to second variations are masked, and by having such a brief test, the day part variations are masked. The delivery of isochronous content requires consistent bandwidth availability second by second, for minutes to hours at a time. When bandwidth logjams occur, even for a second, the content is interrupted, and its value is diminished. And logjams are guaranteed, even for local content, by the extent of sharing of cable modem bandwidth. At present, usage profiles focus on content that is appropriate for bursty bandwidth availability, such that I get to download a page at high speed, then while I'm reading it, the bandwidth is available for many other users -- and each of us has the feel of a high speed connection. The actual average data rate for a cable modem connection is about 10 Kbits per second, if everyone were on line and trying to download at the same moment (which fortunately doesn't happen). A single 40 Mb/s cable modem channel is shared by 200 to 4000 subscribers, and the availability of bandwidth is a statistical gamble. As a result, current "broadband" deployments are not appropriate for widespread use for video content. At 300Kb/s to 1.5Mb/s, only 25 to 125 simultaneous users could be supported, and even then there is not likely to be enough bandwidth from the headend to the Internet backbone.
Contrast this with the deployment of "digital cable". When a channel is assigned to a modulator, the bandwidth to support it is guaranteed, all the way from the signal source to the subscriber. There are two models: Broadcast, where many subscribers view the same digital channel, without the possibility of interaction: and on demand, where each subscriber receives a dedicated bandwidth allocation on a modulator. As a result, in a typical digital headend, there might be one rack of cable modem equipment, but many racks of downstream digital cable modulators (especially when on demand services are offered). A typical DOCSIS cable modem termination system (CMTS) costs $25,000 and provides one modulator, which makes it impractical in size and cost to expand cable modem bandwidth substantially.
The creation of broadband content has been retarded by the relatively slow adoption of broadband. Some have used the chicken vs. egg analogy. The answer, by the way, is "egg". The bird that laid the egg was not a chicken -- close, but not identical. The problem is what is meant by “broadband content”. Most people interpret this as true, interactive video. However, this can’t be deployed with cable modems without a significant change in infrastructure. Right now, effective video delivery uses a completely separate protocol than data or telephony over the cable, and trying to make genuine, synchronized interactivity happen is remarkably difficult – therefore, most “interactive television” ventures over the last decade have fallen by the wayside, even those few with good ideas.
This is why the technology of Advent Networks is so promising. Advent provides dedicated (unshared) downstream data channels for individual subscribers of 5 to 40 Mb/s each, using Internet Protocol (IP), over the Hybrid Fiber Coax (HFC) architecture. Cable operators have spent ten years and $55 billion deploying HFC, and Advent requires no further infrastructure changes.
Advent does this by providing high density, low cost headend modulators, corresponding termination equipment, and a network design capable of supporting all services over the resulting composite bandwidth. Their system (which we invented) is the only practical solution to provide a true broadband connection equivalent to what the press assumes a cable modem connection to be. Further, all services are delivered using IP, (e.g. video, data, telephony, home security, appliance and utility monitoring and control, etc.) which makes interactivity easy. There is someone on every block who can design a decent interactive web site, and video is just a new tool.
2007 update: Everything over IP, including video (IPTV), is going gangbusters, as predicted, with the biggest impacts yet to come. Unfortunately, we must observe a moment of silence for Advent. They bit the dust for three reasons. First, they fell prey to the venture capital belief that the valuation of a technology company is directly proportional to the number of engineers it hires. In truth, the value is maximized at one, or a few, and declines with more cooks and a fixed amount of broth. Second, they refused in spite of repeated prompting and even introductions to participate in the Cablelabs standards process (or any other). They should have set the standard for Docsis 3. Third, the CEO we all really liked, who was a visionary, started thinking with the wrong head. Remember God's two great gifts to man, and the good news and bad? The good: Two great gifts. The bad: Only enough blood to run one at a time. Only a few know that I was behind the blood drive for President Clinton -- It was my intent to fill him up, so that he could remember "Oh, right, I'm the president of the United States!" at critical moments.
Steve Rose is Advent Networks' "Digital Architect". Check out www.advent.net, where you will find no mention of him.